Educational resource retrieval method and device based on cognitive diagnosis and learning path

Abstract

The invention discloses an educational resource retrieval method and device based on cognitive diagnosis and a learning path, and the method comprises the steps: determining a test question-knowledge point matrix and a user-test question matrix according to a knowledge point set, and inputting the matrixes into a cognitive diagnosis model, so that the model determines the comprehensive mastery degree of each user on each knowledge point, and the learning path of each user is obtained. Obtaining weak knowledge points of the user; the test question-knowledge point matrix comprises a video-knowledge point matrix, an exercise-knowledge point matrix and an other resource-knowledge point matrix; the user-test question matrix comprises a user-video matrix, a user-exercise matrix and a user-other resource matrix. Scanning the knowledge point set by using an AprioriAll algorithm, and generating a recommended learning path by improving weight values of weak knowledge points of a user in a learning path planning process; and a plurality of member search engines are introduced to perform resource retrieval on the recommended learning path and integrate retrieval results to obtain a recommendation result, so that the accuracy of educational resource retrieval is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of cognitive diagnosis methods, and in particular relates to an educational resource retrieval method and device based on cognitive diagnosis and learning paths. Background Art

[0002] With the rapid development of Internet technology and the widespread application of computer technology, human beings have entered an era of information explosion. Learners no longer limit their access to knowledge to learning scenarios such as books and classrooms, but instead shift more of their energy to online learning. Then, when learners face a massive amount of educational resources, they often fall into the problems of "information overload" and "resource wandering", which reduces their learning efficiency and interest. Therefore, how to recommend personalized search content to learners to meet their needs for learning and obtaining materials is a research hotspot in the field of education.

[0003] In the online learning scenario, the key to overcoming the two major problems of "learning wandering" and "cognitive overload" is how to help learners select the resources that suit them from the growing learning resources, and organize the learning resources into a coherent learning path according to the learners' learning interests, cognitive levels and the relationship between learning resources, to help learners improve their learning efficiency. Therefore, under the current trend of "Internet + Education", building a personalized learning path recommendation system to adapt to the cognitive differences of different learners has become a hot topic for many scholars to study, but the current learning resource recommendation generally has the following two prominent problems:

[0004] (1) The mainstream way of learning path recommendation is to recommend based entirely on learners’ interests, ignoring learners’ individual weak knowledge points. However, unlike the idea of ​​recommendation based on user preferences in other fields, the real significance of recommendation systems in the field of smart education should be to help learners quickly identify and make up for their individual weak knowledge points. Mainstream methods cannot help learners find gaps and effectively improve their learning efficiency.

[0005] (2) The mainstream method of diagnosing weak knowledge points is to judge the learner's mastery of knowledge points based on exercise data. However, with the spread of streaming video, the mainstream method only considers the exercise data and cannot obtain an accurate mastery of knowledge points.

[0006] However, in today's education field, existing cognitive diagnosis methods cannot accurately obtain learners' weak knowledge points. The traditional DINA model (Deterministic Inputs, Noisy "And" gate model, a typical discrete cognitive diagnosis model) is used for cognitive diagnosis. Due to the use of a two-level scoring mechanism, it is impossible to discretize the learner's mastery of knowledge points into a continuous interval of [0, 1]. In addition, traditional learning path recommendation algorithms often have cold start problems and do not take into account the learner's own learning situation.

[0007] Therefore, how to accurately obtain learners' weak knowledge points to generate appropriate recommended learning paths for learners and obtain more accurate educational resource retrieval results has become an important issue. Summary of the invention

[0008] In order to solve the above problems existing in the prior art, the present invention provides an educational resource retrieval method and device based on cognitive diagnosis and learning path.

[0009] The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0010] In a first aspect, the present invention provides an educational resource retrieval method based on cognitive diagnosis and learning paths, the educational resource retrieval method comprising:

[0011] Determine the test question-knowledge point matrix and the user-test question matrix according to the knowledge point set; the test question-knowledge point matrix includes a video-knowledge point matrix, an exercise-knowledge point matrix and an other resource-knowledge point matrix; the user-test question matrix includes a user-video matrix, a user-exercise matrix and a user-other resource matrix; the test question-knowledge point matrix is ​​used to indicate the degree of association between the test question and the knowledge point; the user-test question matrix is ​​used to indicate the user's answer to the test question;

[0012] The test question-knowledge point matrix and the user-test question matrix are input into a cognitive diagnosis model, so that the cognitive diagnosis model can obtain the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point; the user's weak knowledge points are knowledge points whose comprehensive mastery is less than a preset threshold;

[0013] Scan the knowledge point set using the AprioriAll algorithm, combine the user's weak knowledge points, and generate a recommended learning path by increasing the weight value of the user's weak knowledge points during the learning path planning process;

[0014] A plurality of member search engines are introduced to perform resource retrieval on the recommended learning path, and the retrieval results of each member search engine are integrated to obtain a recommendation result.

[0015] Optionally, the cognitive diagnosis model includes a VFuzzy-CDF model, a CFuzzy-CDF model and a Fuzzy-CDF model, and the test question-knowledge point matrix and the user-test question matrix are input into the cognitive diagnosis model, so that the cognitive diagnosis model obtains the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point, including:

[0016] Inputting the video-knowledge point matrix and the user-video matrix into a VFuzzy-CDF model, so that the VFuzzy-CDF model obtains the user's first weak knowledge point by judging the comprehensive mastery of each knowledge point by each user;

[0017] Input the other resource-knowledge point matrix and the user-other resource matrix into the CFuzzy-CDF model, so that the CFuzzy-CDF model can obtain the second weakest knowledge point of each user by judging the comprehensive mastery of each knowledge point by each user;

[0018] Inputting the exercise-knowledge point matrix and the user-exercise matrix into a fuzzy-CDF model, so that the fuzzy-CDF model can obtain the third weakest knowledge point of each user by judging the comprehensive mastery of each knowledge point by each user;

[0019] The user's first weak knowledge point, the user's second weak knowledge point and the user's third weak knowledge point are integrated to obtain the user's weak knowledge point.

[0020] Optionally, the test question-knowledge point matrix and the user-test question matrix are input into a cognitive diagnosis model, so that the cognitive diagnosis model obtains the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point in the following manner:

[0021] According to the test question-knowledge point matrix, a fuzzy attribute weight of each test question to each knowledge point is obtained; the fuzzy attribute weight is used to indicate the relevance of each test question to each knowledge point;

[0022] According to the fuzzy attribute weight of each test question for each knowledge point and each user's mastery of each knowledge point, the mastery degree of each user on the required knowledge point in each test question is obtained;

[0023] Based on each user's mastery of the knowledge points required in each test question, the user's error rate and guessing rate are introduced to obtain each user's actual score rate;

[0024] The comprehensive mastery of each knowledge point by the user is determined by combining the actual score rate of each user and the probability of the user's correct answer to each question in the question-knowledge point matrix, and the knowledge points with a comprehensive mastery level less than a preset threshold are regarded as the user's weak knowledge points.

[0025] Optionally, multiple member search engines are introduced to perform resource retrieval on the recommended learning path, and the retrieval results of each member search engine are integrated to obtain recommendation results, including:

[0026] Introducing multiple member search engines to perform resource retrieval on the recommended learning path, and integrating the retrieval results of each member search engine to obtain an initial retrieval result;

[0027] After removing invalid search results from the initial search results, the remaining search results are sorted to obtain recommended results.

[0028] Optionally, after removing invalid search results from the initial search results, the remaining search results are sorted to obtain recommended results, including:

[0029] After removing invalid search results from the initial search results, the remaining search results are sorted in combination with the accuracy, recall and nDCG of each member search engine to obtain recommendation results.

[0030] In a second aspect, the present invention provides an educational resource retrieval device based on cognitive diagnosis and learning paths, the educational resource retrieval device comprising:

[0031] A matrix determination module is used to determine a test question-knowledge point matrix and a user-test question matrix according to a knowledge point set; the test question-knowledge point matrix includes a video-knowledge point matrix, an exercise-knowledge point matrix and an other resource-knowledge point matrix; the user-test question matrix includes a user-video matrix, a user-exercise matrix and a user-other resource matrix; the test question-knowledge point matrix is ​​used to indicate the degree of association between the test question and the knowledge point; the user-test question matrix is ​​used to indicate the user's answer to the test question;

[0032] A user weak knowledge point determination module is used to input the test question-knowledge point matrix and the user-test question matrix into a cognitive diagnosis model, so that the cognitive diagnosis model can obtain the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point; the user's weak knowledge points are knowledge points whose comprehensive mastery is less than a preset threshold;

[0033] A recommended learning path generation module, used to scan the knowledge point set using the AprioriAll algorithm, and generate a recommended learning path by increasing the weight value of the user's weak knowledge points in the learning path planning process in combination with the user's weak knowledge points;

[0034] The search result integration module is used to introduce multiple member search engines to perform resource search on the recommended learning path, and integrate the search results of each member search engine to obtain a recommendation result.

[0035] Optionally, the cognitive diagnosis model includes a VFuzzy-CDF model, a CFuzzy-CDF model and a Fuzzy-CDF model, and the user weak knowledge point determination module is specifically used to input the video-knowledge point matrix and the user-video matrix into the VFuzzy-CDF model, so that the VFuzzy-CDF model obtains the user's first weak knowledge point by judging the comprehensive mastery of each user on each knowledge point; input the other resource-knowledge point matrix and the user-other resource matrix into the CFuzzy-CDF model, so that the CFuzzy-CDF model obtains the user's second weak knowledge point by judging the comprehensive mastery of each user on each knowledge point; input the exercise-knowledge point matrix and the user-exercise matrix into the Fuzzy-CDF model, so that the Fuzzy-CDF model obtains the user's third weak knowledge point by judging the comprehensive mastery of each user on each knowledge point; and integrate the user's first weak knowledge point, the user's second weak knowledge point and the user's third weak knowledge point to obtain the user's weak knowledge point.

[0036] Optionally, the user weak knowledge point determination module inputs the test question-knowledge point matrix and the user-test question matrix into a cognitive diagnosis model, so that the cognitive diagnosis model obtains the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point in the following manner:

[0037] According to the test question-knowledge point matrix, a fuzzy attribute weight of each test question to each knowledge point is obtained; the fuzzy attribute weight is used to indicate the relevance of each test question to each knowledge point;

[0038] According to the fuzzy attribute weight of each test question for each knowledge point and each user's mastery of each knowledge point, the mastery degree of each user on the required knowledge point in each test question is obtained;

[0039] Based on each user's mastery of the knowledge points required in each test question, the user's error rate and guessing rate are introduced to obtain each user's actual score rate;

[0040] The comprehensive mastery of each knowledge point by the user is determined by combining the actual score rate of each user and the probability of the user's correct answer to each question in the user-question matrix, and the knowledge points with a comprehensive mastery level less than a preset threshold are regarded as the user's weak knowledge points.

[0041] Optionally, the retrieval result integration module is specifically used to introduce multiple member search engines to perform resource retrieval on the recommended learning path, and integrate the retrieval results under each member search engine to obtain the initial retrieval result; after removing invalid retrieval results from the initial retrieval result, the remaining retrieval results are sorted to obtain the recommended result.

[0042] Optionally, the search result integration module removes invalid search results from the initial search results and sorts the remaining search results to obtain recommendation results, including:

[0043] After removing invalid search results from the initial search results, the remaining search results are sorted in combination with the accuracy, recall and nDCG of each member search engine to obtain recommendation results.

[0044] The present invention provides an educational resource retrieval method based on cognitive diagnosis and learning path, which determines the test question-knowledge point matrix and the user-test question matrix according to the knowledge point set, wherein the test question-knowledge point matrix includes a video-knowledge point matrix, an exercise-knowledge point matrix and an other resource-knowledge point matrix, and the user-test question matrix includes a user-video matrix, a user-exercise matrix and a user-other resource matrix; the test question-knowledge point matrix and the user-test question matrix are input into the cognitive diagnosis model, and the user's weak knowledge points are diagnosed from multiple dimensions of video, exercises and other resources by using the cognitive diagnosis model, thereby improving the accuracy of the judgment of the user's weak knowledge points. In the process of learning path planning using the AprioriAll algorithm, by increasing the weight value of the user's weak knowledge points, it is ensured that the user's weak knowledge points are paid more attention to in generating the recommended learning path, and the accuracy of the educational resource retrieval results is guaranteed. Moreover, in the present invention, by introducing multiple member search engines to retrieve resources for the recommended learning path, the shortcomings of low coverage and insufficient personalization in the traditional retrieval method are solved.

[0045] The present invention will be further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flowchart of an educational resource retrieval method based on cognitive diagnosis and learning path provided by an embodiment of the present invention;

[0047] Figure 2 It is a schematic diagram of a scenario using a data structure as a knowledge point set provided by an embodiment of the present invention;

[0048] Figure 3 is a schematic diagram of the structure of a cognitive diagnosis model provided by an embodiment of the present invention;

[0049] Figure 4It is a schematic diagram of a process of scanning a knowledge point set using the AprioriAll algorithm provided in an embodiment of the present invention;

[0050] Figure 5 It is a schematic diagram of a resource search process provided by an embodiment of the present invention;

[0051] Figure 6 An educational resource retrieval device based on cognitive diagnosis and learning paths is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.

[0053] In order to solve the problem that the existing cognitive diagnosis method cannot accurately obtain the learner's weak knowledge points, the embodiment of the present invention provides an educational resource retrieval method based on cognitive diagnosis and learning path, see Figure 1 , Figure 1 : is a flowchart of an educational resource retrieval method based on cognitive diagnosis and learning path provided by an embodiment of the present invention, which specifically includes the following steps:

[0054] Step S101, determine the test question-knowledge point matrix and the user-test question matrix based on the knowledge point set; the test question-knowledge point matrix includes the video-knowledge point matrix, the exercise question-knowledge point matrix and the other resource-knowledge point matrix; the user-test question matrix includes the user-video matrix, the user-exercise matrix and the user-other resource matrix; the test question-knowledge point matrix is ​​used to indicate the degree of correlation between the test question and the knowledge point; the user-test question matrix is ​​used to indicate the user's answer to the test question.

[0055] In an embodiment of the present invention, determining the question-knowledge point matrix and the user-question matrix according to the knowledge point set includes extracting the question-knowledge point matrix and the user-question matrix from a preset data set according to the knowledge point set.

[0056] The preset data set provided in the embodiment of the present invention is obtained based on the MOOCCubeX data set. The MOOCCubeX data set is an open data warehouse for researchers of natural language processing, knowledge graphs, data mining, etc. related to large-scale online education. It contains JSON data such as learners, that is, users' answers to exercises, video viewing, and learners' comments, where JSON is an open standard file format and data exchange format.

[0057] Extracting the test question-knowledge point matrix and user-test question matrix from the data set according to the knowledge point set specifically includes extracting the video-knowledge point matrix, exercise-knowledge point matrix, other resource-knowledge point matrix, user-video matrix, user-exercise matrix and user-other resource matrix from the data set according to the knowledge point set. Among them, other resources can be comments of users under the relevant information of a certain knowledge point.

[0058] The video-knowledge point matrix represents the proportion of relevant information of knowledge points in the video, that is, the degree of association between knowledge points and videos. For example, in the video-knowledge point matrix, rows represent videos, columns represent knowledge points, and the value of the a-th row and the b-th column is 0.3, then the proportion of the b-th knowledge point in the a-th video is 0.3.

[0059] The exercise-knowledge point matrix represents the proportion of relevant information of knowledge points in the exercises, that is, the degree of association between knowledge points and exercises. For example, in the exercise-knowledge point matrix, rows represent exercises, columns represent knowledge points, and the value of the cth row and dth column is 0.5, then the proportion of the dth knowledge point in the cth exercise is 0.5.

[0060] The other resources-knowledge point matrix indicates the proportion of relevant information of knowledge points in other resources, that is, the degree of association between knowledge points and other resources. For example, in the other resources-knowledge point matrix, rows represent other resources, columns represent knowledge points, and the value of the e-th row and the f-th column is 0.7, then the proportion of the f-th knowledge point in the e-th other resource is 0.7.

[0061] In the embodiment of the present invention, the user-video matrix represents the user's viewing time of the video; the user-exercise matrix represents the user's answers to the exercises; and the user-other resource matrix represents the user's mastery of other resources.

[0062] Based on the video-knowledge point matrix and the user-video matrix, the user's learning degree of the video can be represented by the user's viewing time of the video, and finally the user's mastery of the relevant knowledge points in the video can be obtained.

[0063] Based on the exercise-knowledge point matrix and the user-exercise matrix, the user's mastery of a knowledge point can be represented by the user's correct rate of related exercises for a certain knowledge point.

[0064] Based on the other resources-knowledge point matrix and the user-other resources matrix, the user's mastery of a knowledge point can be represented by the user's activeness in commenting on the content related to a knowledge point.

[0065] See also Figure 2 , Figure 2: is a schematic diagram of a scenario with a data structure as a knowledge point set provided by an embodiment of the present invention. Taking the data structure as a knowledge point set as an example, the knowledge point set may specifically include a tree, a graph, a layer-order traversal algorithm, a sorting algorithm, a sequence list, a linked list, a greedy algorithm, a backtracking algorithm, a depth search traversal, and a breadth search traversal, etc. Exercises e1, e2, e3, and e4 represent Exercises 1, Exercises 2, Exercises 3, and Exercises 4. The answer status is used to indicate whether the user correctly answers the exercise. Among them, the symbol "√" indicates that the user correctly answers the exercise, and the symbol "×" indicates that the user does not correctly answer the exercise. Exercises v1, v2, v3, and v4 represent Video 1, Video 2, Video 3, and Video 4. Video playback is used to indicate whether the user has finished playing the video. Among them, the symbol "√" indicates that the user has finished playing the video, and the symbol "×" indicates that the user has not finished playing the video. When the other resources are comments, Exercises o1, o2, o3, and o4 represent Comment Status 1, Comment Status 2, Comment Status 3, and Comment Status 4. The comment status is used to indicate whether the user has made a comment. The symbol "√" indicates that the user has made a comment, and the symbol "×" indicates that the user has not made a comment.

[0066] In the embodiment of the present invention, the Q matrix can be used to represent the test question-knowledge point matrix, where qvk=1, indicating that the test question Jv contains the knowledge point Sk, and qvk=0, indicating that the test question Jv does not contain the knowledge point Sk. The V matrix represents the user-test question matrix, where ηuv=1, indicating that the user Pu answered the test question Jv correctly, and ηuv=0, indicating that the user Pu did not answer the test question Jv correctly.

[0067] Step S102, input the test question-knowledge point matrix and the user-test question matrix into the cognitive diagnosis model, so that the cognitive diagnosis model can obtain the user's weak knowledge points by judging each user's comprehensive mastery of each knowledge point; the user's weak knowledge points are knowledge points whose comprehensive mastery is less than a preset threshold.

[0068] In an embodiment of the present invention, by taking each user's test question-knowledge point matrix and user-test question matrix as input, the cognitive diagnosis model can quantify the mastery level of each user on each knowledge point into a continuous value, and finally output the comprehensive mastery level of each user on each knowledge point.

[0069] In the embodiments of the present invention, see Figure 3 , Figure 3 : is a schematic diagram of the structure of the cognitive diagnosis model provided by an embodiment of the present invention. The test question-knowledge point matrix and the user-test question matrix are input into the cognitive diagnosis model, so that the cognitive diagnosis model obtains the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point in the following manner:

[0070] a) Calculate the fuzzy attribute weights of the test questions for the knowledge points.

[0071] The fuzzy attribute weights of each test question to each knowledge point are obtained according to the test question-knowledge point matrix; the fuzzy attribute weights are used to indicate the relevance of each test question to each knowledge point.

[0072] In the embodiment of the present invention, the fuzzy attribute weight α of the jth test question to the kth knowledge point jk It can be understood as the correlation between the jth test question and the kth knowledge point. When the correlation between the jth test question and the kth knowledge point is high, the fuzzy attribute weight of the jth test question to the kth knowledge point is high. On the contrary, when the correlation between the jth test question and the kth knowledge point is low, the fuzzy attribute weight of the jth test question to the kth knowledge point is low. Where j = 1, 2, ... M, k = 1, 2, ..., K, M represents the total number of test questions, and K represents the total number of knowledge points.

[0073] Specifically, if α jk =1, indicating that the jth test question is completely related to the kth knowledge point, that is, the jth test question directly contains the kth knowledge point. jk =0, indicating that the jth test question is completely irrelevant to the kth knowledge point, that is, the jth test question does not contain the kth knowledge point at all. jk The value range can be between [0,1].

[0074] In an embodiment of the present invention, the fuzzy attribute weights of test questions to knowledge points are determined based on the positive correlation weights of each test question and each knowledge point, the negative correlation weights of each test question and each knowledge point, and the difficulty of the test question.

[0075] Specifically, α jk According to the positive correlation weight a between the jth test question and the kth knowledge point jk , the negative correlation weight b between the jth test question and the kth knowledge point jk And the difficulty of the jth test question determines θ j .

[0076] Among them, the positive correlation weight a jk Used to indicate the importance of the kth knowledge point in the jth test question. Negative correlation weight b jk Used to indicate the influence of interference factors or non-target attributes of the kth knowledge point in the jth test question.

[0077] b) Determine the user's mastery of the knowledge points required in the test questions.

[0078] According to the fuzzy attribute weight of each test question for each knowledge point and the mastery degree of each user on each knowledge point, the mastery degree of each user on the knowledge points required in each test question is obtained.

[0079] In the embodiment of the present invention, the degree to which the i-th user has mastered the knowledge points required in the j-th test question can be expressed as ηji Indicates. ji It is a broad ability assessment indicator, usually used to reflect whether the user has a sufficient grasp of the knowledge points required in the test questions. ji The larger it is, the more fully the i-th user has mastered the knowledge points required for the j-th test question.

[0080] η ji The specific calculation method is:

[0081]

[0082] Among them, η ji According to q ik and q jk OK, q ik Indicates the mastery of the kth knowledge point by the ith user, and its value is in the range of [0, 1]. jk Represents the fuzzy attribute weight of the jth test question to the kth knowledge point. i=1,2,…,N o oeLh s , N o represents the set of knowledge points, N s Represents a subset of the knowledge point set.

[0083] c) Determine the user's actual answer results to the test questions.

[0084] Based on each user's mastery of the knowledge points required in each test question, the user's error rate and guessing rate are introduced to obtain each user's actual score rate.

[0085] In the embodiment of the present invention, the actual answer result of the i-th user to the j-th test question can be obtained by the actual score rate R ji To represent, where R ji The value of is 0 or 1.

[0086] In the embodiment of the present invention, the user's mastery of the knowledge points required in the test questions, the error rate s and the guessing rate g can be introduced to model the user's answering of the test questions in a real state. Among them, the error rate s represents the probability that the user knows the knowledge point but answers the test question incorrectly, and the guessing rate g represents the probability that the user does not know the knowledge point but answers the test question correctly. By introducing the error rate s and the guessing rate g, the user's mastery of the knowledge points can be further restored, and the actual answer result of the user to the test question can be obtained.

[0087] d) Determine the user's comprehensive mastery of each knowledge point by combining each user's actual score rate and the probability of the user's correct answer to each question in the user-question matrix.

[0088] In the embodiment of the present invention, the probability μ of the correct answer of the i-th user to the j-th question isij and R ji The comprehensive mastery level of each user on each knowledge point can be determined.

[0089] The probability μ that the i-th user correctly answers the j-th question ij is a probability value that can be directly used to reflect the possibility that the i-th user correctly answers the j-th question, μ ij The calculation method is as follows:

[0090]

[0091] in, It means that if q jk =1, the expression is Indicates that the jth test question contains the kth knowledge point, so it is necessary to consider the user's mastery of the kth knowledge point. If q jk =0, Indicates that the jth test question does not contain the kth knowledge point. ik Indicates the mastery of the kth knowledge point by the i-th user. jk Indicates the possibility that the j-th test question contains the k-th knowledge point, and the value is 0 or 1. When the value is 1, the j-th test question contains the k-th knowledge point; otherwise, it does not contain it.

[0092] In the embodiment of the present invention, the knowledge points whose comprehensive mastery degree is less than a preset threshold are regarded as the user's weak knowledge points.

[0093] The specific preset threshold value can be set by the technical staff according to the requirements and is not limited here.

[0094] Step S103, using the AprioriAll algorithm to scan the knowledge point set, combining the user's weak knowledge points, and generating a recommended learning path by increasing the weight value of the user's weak knowledge points during the learning path planning process.

[0095] In the embodiment of the present invention, the AprioriAll algorithm is an association rule mining algorithm, which can be used to discover frequent set items in a knowledge point set.

[0096] See also Figure 4 , Figure 4It is a flow chart of scanning a knowledge point set using the AprioriAll algorithm provided by an embodiment of the present invention, determining the information of the learner, that is, the user's information, and determining the learning goal using the AprioriAll algorithm. In the AprioriAll algorithm, a minimum support threshold min_support is set, and min_support can be used to screen the support threshold of the frequent item set. The comprehensive mastery degree of each knowledge point obtained by the cognitive diagnosis model is used to traverse the knowledge point set based on the depth first traversal based on the weak knowledge point or the breadth first traversal based on the weak knowledge point to generate a frequent 1 item set. By scanning the knowledge point set, the number of occurrences of each individual item is calculated, and those items with support greater than or equal to min_support are retained to form a frequent 1 item set. Then, when generating a candidate t item set, a frequent (t-1) item set is used to generate a candidate t item set. If any subset of a candidate item set is not frequent, then the candidate item set cannot be frequent. The candidate item set with support greater than or equal to min_support is retained to form a frequent t item set. When generating frequent item sets, a higher weight is given to the learning content of the user's weak knowledge point obtained by the cognitive diagnosis model to improve its priority in recommendation. Generate association rules in frequent item sets, set the minimum confidence threshold, filter out learning content combinations with strong correlation, pay special attention to content related to weak knowledge points, ensure that these users' weak knowledge points are included in the recommendation, and give priority to recommending learning content that can make up for these users' weak knowledge points according to the users' weak knowledge points, and recommend learning paths to help users improve relevant knowledge. Evaluate the effectiveness of recommendations through user feedback and learning effects, and adjust the support and confidence thresholds when necessary to optimize the recommendation results.

[0097] Specifically, the support calculation method includes:

[0098]

[0099] Among them, Support(x) represents the support of item set x; support represents the frequency of occurrence of a certain item set x in the entire knowledge point data set. conunt(x) represents the number of knowledge point records containing item set x; N represents the total number of records in the knowledge point data set.

[0100]

[0101] Among them, confidence(x→y) indicates the probability that item set y is also included when item set x is included, x→y: indicates an association rule; support(x∪y): indicates the support of both item set x and item set y.

[0102] Step S104, introducing multiple member search engines to perform resource search on the recommended learning path, and integrating the search results of each member search engine to obtain a recommendation result.

[0103] In the embodiment of the present invention, in order to address the shortcomings of the traditional education search system in terms of insufficient coverage accuracy, multiple member search engines are introduced to integrate the search results to obtain recommendation results. Since the present invention is aimed at the field of education, when selecting a member search engine, multiple dimensions such as educational text information and educational videos should be considered.

[0104] In an embodiment of the present invention, a test question-knowledge point matrix and a user-test question matrix are determined according to a knowledge point set, wherein the test question-knowledge point matrix includes a video-knowledge point matrix, an exercise-knowledge point matrix and an other resource-knowledge point matrix, and the user-test question matrix includes a user-video matrix, a user-exercise matrix and a user-other resource matrix; the test question-knowledge point matrix and the user-test question matrix are input into a cognitive diagnosis model, and the user's weak knowledge points are diagnosed from multiple dimensions of video, exercises and other resources by using the cognitive diagnosis model, thereby improving the accuracy of the judgment of the user's weak knowledge points. In the process of learning path planning using the AprioriAll algorithm, by increasing the weight value of the user's weak knowledge points, it is ensured that more attention is paid to the user's weak knowledge points in generating the recommended learning path, thereby ensuring the accuracy of the educational resource retrieval results. Moreover, in an embodiment of the present invention, by introducing multiple member search engines to retrieve resources for the recommended learning path, the shortcomings of low coverage and insufficient personalization in the traditional retrieval method are solved.

[0105] In one implementation, the cognitive diagnosis model includes a VFuzzy-CDF model, a CFuzzy-CDF model, and a Fuzzy-CDF model. The test question-knowledge point matrix and the user-test question matrix are input into the cognitive diagnosis model, so that the cognitive diagnosis model can determine the comprehensive mastery of each user on each knowledge point and obtain the user's weak knowledge points, including:

[0106] The video-knowledge point matrix and the user-video matrix are input into the VFuzzy-CDF model, so that the VFuzzy-CDF model can obtain the user's first weak knowledge point by judging the comprehensive mastery of each knowledge point by each user;

[0107] Input the other resources-knowledge point matrix and the user-other resources matrix into the CFuzzy-CDF model, so that the CFuzzy-CDF model can obtain the second weakest knowledge point of each user by judging the comprehensive mastery of each knowledge point of each user;

[0108] Input the exercise-knowledge point matrix and the user-exercise matrix into the Fuzzy-CDF model, so that the Fuzzy-CDF model can obtain the third weakest knowledge point of each user by judging the comprehensive mastery of each knowledge point of each user;

[0109] The user's first weak knowledge point, the user's second weak knowledge point and the user's third weak knowledge point are integrated to obtain the user's weak knowledge point.

[0110] In the embodiment of the present invention, VFuzzy-CDF and CFuzzy-CDF are improved based on Fuzzy-CDF, which is a knowledge tracking model. The existing Fuzzy-CDF can only be used to judge the user's mastery of knowledge points based on exercises, and its judgment basis is obtained by extracting the user-exercise matrix and exercise-knowledge point matrix from the upstream knowledge point set.

[0111] In the VFuzzy-CDF provided in the embodiment of the present invention, the extracted matrix is ​​improved, the video-knowledge point matrix and the user-video matrix are used as the input matrices of VFuzzy-CDF, and the judgment basis of the model is improved to the user's learning degree of the video and the proportion of relevant knowledge points in the video.

[0112] In the CFuzzy-CDF provided in the embodiment of the present invention, other resources are considered. Specifically, other resources can be the comment activity of users posting comments in relevant comment areas. The input matrix of the model is improved to other resource-knowledge point matrix and user-other resource matrix.

[0113] For example, in the VFuzzy-CDF model, there are U users, V videos, and K knowledge points. The video-knowledge point matrix and the user-video matrix are extracted, and the user's learning degree of the video is represented by the user's viewing time of the video, combined with the proportion of relevant knowledge points in the video, and finally the user's mastery of the knowledge points in the video is obtained. The error rate and guessing rate are introduced to represent the probability that the user has mastered the knowledge point but has not finished watching the video and the probability that the user has not mastered the knowledge point but has finished watching the video, respectively, to simulate the situation of users watching videos in real state.

[0114] For example, in the CFuzzy-CDF model, there are U users, C comments, and K knowledge points. The other resource-knowledge point matrix and the user-other resource matrix are extracted, and the user's mastery of a certain knowledge point is represented by the user's activeness in commenting on the knowledge point, and the error rate and guess rate are also introduced to represent the situation of users posting comments in real state. Among them, other resources can refer to comments.

[0115] Specifically, Fuzzy-CDF is used to obtain the user's first weak knowledge point by judging the user's comprehensive ability matching to the corresponding knowledge points in the exercises; VFuzzy-CDF is used to obtain the user's second weak knowledge point by judging the user's comprehensive ability matching to the corresponding knowledge points in the video; CFuzzy-CDF is used to obtain the user's third weak knowledge point by judging the user's comprehensive ability matching to the corresponding knowledge points in the comments.

[0116] In the embodiment of the present invention, the user's weak knowledge points may be obtained by fusing the user's first weak knowledge points, the user's second weak knowledge points, and the user's third weak knowledge points.

[0117] In one implementation, multiple member search engines are introduced to perform resource retrieval on the recommended learning path, and the search results of each member search engine are integrated to obtain recommendation results, including:

[0118] Introduce multiple member search engines to search for resources on the recommended learning path, and integrate the search results of each member search engine to obtain the initial search results;

[0119] After removing invalid search results from the initial search results, the remaining search results are sorted to obtain recommended results.

[0120] See also Figure 5 , Figure 5 The resource search process diagram provided by the embodiment of the present invention is shown in FIG. The query interface introduces multiple member search engines to search for resources on the recommended learning path, and integrates the search results of each member search engine to obtain the initial search results. The member search engines are egine 1, egine 2, etc., egine n.

[0121] Then, invalid search results are removed from the search result set, i.e., the initial search results, by determining whether the URLs (uniform resource location) are the same, determining the content of the abstract and title to achieve deduplication, and removing dead links by pinging the URL.

[0122] Specifically, invalid link checks and duplicate information removal are performed. According to statistics, when users browse the query results, they often only look at the first 20-30 results, so the feasibility of the link can be checked for each query result. By sending a query request to the link, if the response code returns 404 or 500, the link is judged as an invalid link and deleted directly. The initial search results come from various member search engines, so there are often many duplicate results in the initial search results. These duplicate results can be used by multiple underlying retrieval systems, which also represents the importance of the record. However, if they are not processed, it will further cause users to form "information wandering" and increase the burden on users. Therefore, the retrieval system needs to judge the repeatability of the query results by comparing the URLs, titles, and abstracts of the two query results. If the URLs of the two query results are exactly the same, they are duplicate results. If the URLs are different, the document titles and abstracts are the same, they are also marked as duplicate results.

[0123] In one implementation, after removing invalid search results from the initial search results, the remaining search results are sorted to obtain recommended results, including:

[0124] After removing invalid search results from the initial search results, the remaining search results are ranked based on the precision, recall and nDCG of each member search engine to obtain the recommended results.

[0125] In the embodiment of the present invention, the accuracy is calculated as follows:

[0126]

[0127] The recall rate is calculated as:

[0128]

[0129] In the embodiment of the present invention, accuracy and recall are two commonly used indicators in information retrieval research. Accuracy refers to the proportion of correct duplicate web pages among the duplicate web pages detected by the search engine, and recall refers to the ratio of the total number of duplicate web pages found by the search engine among all the duplicate web pages.

[0130] nDCG (Normalized Discounted Cumulative Gain) is a retrieval metric designed for search engine evaluation. Introducing weak knowledge points and successor knowledge points, among all the remaining retrieval results, the first P retrieval results are usually recommended. Therefore, the nDCG of the first P retrieval results, that is, nDCG p The calculation method is:

[0131]

[0132] Where P represents the length of the remaining search result list, usually the first P search results; rel represents the relevance score of the pth search result; p represents the ranking position of the search result; IDCG P The ideal DCG value for the first P results is the DCG value obtained by sorting the results in descending order of relevance.

[0133] After sorting the remaining search results by combining the accuracy, recall and nDCG of each member search engine, the recommended results are obtained, displayed and returned to the user.

[0134] Based on the same inventive concept, the embodiment of the present invention also provides an educational resource retrieval device based on cognitive diagnosis and learning path, see Figure 6 , Figure 6 An educational resource retrieval device based on cognitive diagnosis and learning path is provided in an embodiment of the present invention, and the educational resource retrieval device includes:

[0135] The matrix determination module 601 is used to determine the test question-knowledge point matrix and the user-test question matrix according to the knowledge point set; the test question-knowledge point matrix includes a video-knowledge point matrix, an exercise-knowledge point matrix and an other resource-knowledge point matrix; the user-test question matrix includes a user-video matrix, a user-exercise matrix and a user-other resource matrix; the test question-knowledge point matrix is ​​used to indicate the degree of association between the test question and the knowledge point; the user-test question matrix is ​​used to indicate the user's answer to the test question;

[0136] The user weak knowledge point determination module 602 is used to input the test question-knowledge point matrix and the user-test question matrix into a cognitive diagnosis model, so that the cognitive diagnosis model can obtain the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point; the user's weak knowledge points are knowledge points whose comprehensive mastery is less than a preset threshold;

[0137] The recommended learning path generation module 603 is used to scan the knowledge point set using the AprioriAll algorithm, and generate a recommended learning path by increasing the weight value of the user's weak knowledge points in the learning path planning process in combination with the user's weak knowledge points;

[0138] The search result integration module 604 is used to introduce multiple member search engines to perform resource search on the recommended learning path, and integrate the search results of each member search engine to obtain a recommendation result.

[0139] In an embodiment of the present invention, a test question-knowledge point matrix and a user-test question matrix are determined according to a knowledge point set, wherein the test question-knowledge point matrix includes a video-knowledge point matrix, an exercise-knowledge point matrix and an other resource-knowledge point matrix, and the user-test question matrix includes a user-video matrix, a user-exercise matrix and a user-other resource matrix; the test question-knowledge point matrix and the user-test question matrix are input into a cognitive diagnosis model, and the user's weak knowledge points are diagnosed from multiple dimensions of video, exercises and other resources by using the cognitive diagnosis model, thereby improving the accuracy of the judgment of the user's weak knowledge points. In the process of learning path planning using the AprioriAll algorithm, by increasing the weight value of the user's weak knowledge points, it is ensured that more attention is paid to the user's weak knowledge points in generating the recommended learning path, thereby ensuring the accuracy of the educational resource retrieval results. Moreover, in an embodiment of the present invention, by introducing multiple member search engines to retrieve resources for the recommended learning path, the shortcomings of low coverage and insufficient personalization in the traditional retrieval method are solved.

[0140] Optionally, the cognitive diagnosis model includes a VFuzzy-CDF model, a CFuzzy-CDF model and a Fuzzy-CDF model, and the user weak knowledge point determination module is specifically used to input the video-knowledge point matrix and the user-video matrix into the VFuzzy-CDF model, so that the VFuzzy-CDF model obtains the user's first weak knowledge point by judging the comprehensive mastery of each user on each knowledge point; input the other resource-knowledge point matrix and the user-other resource matrix into the CFuzzy-CDF model, so that the CFuzzy-CDF model obtains the user's second weak knowledge point by judging the comprehensive mastery of each user on each knowledge point; input the exercise-knowledge point matrix and the user-exercise matrix into the Fuzzy-CDF model, so that the Fuzzy-CDF model obtains the user's third weak knowledge point by judging the comprehensive mastery of each user on each knowledge point; and integrate the user's first weak knowledge point, the user's second weak knowledge point and the user's third weak knowledge point to obtain the user's weak knowledge point.

[0141] Optionally, the user weak knowledge point determination module inputs the test question-knowledge point matrix and the user-test question matrix into a cognitive diagnosis model, so that the cognitive diagnosis model obtains the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point in the following manner:

[0142] According to the test question-knowledge point matrix, a fuzzy attribute weight of each test question to each knowledge point is obtained; the fuzzy attribute weight is used to indicate the relevance of each test question to each knowledge point;

[0143] According to the fuzzy attribute weight of each test question for each knowledge point and each user's mastery of each knowledge point, the mastery degree of each user on the required knowledge point in each test question is obtained;

[0144] Based on each user's mastery of the knowledge points required in each test question, the user's error rate and guessing rate are introduced to obtain each user's actual score rate;

[0145] The comprehensive mastery of each knowledge point by the user is determined by combining the actual score rate of each user and the probability of the user's correct answer to each question in the user-question matrix, and the knowledge points with a comprehensive mastery level less than a preset threshold are regarded as the user's weak knowledge points.

[0146] Optionally, the retrieval result integration module is specifically used to introduce multiple member search engines to perform resource retrieval on the recommended learning path, and integrate the retrieval results under each member search engine to obtain the initial retrieval result; after removing invalid retrieval results from the initial retrieval result, the remaining retrieval results are sorted to obtain the recommended result.

[0147] Optionally, the search result integration module removes invalid search results from the initial search results and sorts the remaining search results to obtain recommendation results, including:

[0148] After removing invalid search results from the initial search results, the remaining search results are sorted in combination with the accuracy, recall and nDCG of each member search engine to obtain recommendation results.

[0149] It should be noted that the terms "first", "second", etc. are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention.

[0150] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification.

[0151] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art can understand and implement other changes to the disclosed embodiments by viewing the drawings and the disclosed content. In the description of the present invention, the term "comprising" does not exclude other components or steps, "one" or "an" does not exclude multiple situations, and "multiple" means two or more, unless otherwise clearly and specifically limited. In addition, certain measures are recorded in different embodiments, but this does not mean that these measures cannot be combined to produce good results.

[0152] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. An educational resource retrieval method based on cognitive diagnosis and learning path, characterized in that: The educational resource retrieval method comprises: Determine the test question-knowledge point matrix and the user-test question matrix according to the knowledge point set; the test question-knowledge point matrix includes a video-knowledge point matrix, an exercise-knowledge point matrix and an other resource-knowledge point matrix; the user-test question matrix includes a user-video matrix, a user-exercise matrix and a user-other resource matrix; the test question-knowledge point matrix is ​​used to indicate the degree of association between the test question and the knowledge point; the user-test question matrix is ​​used to indicate the user's answer to the test question; The test question-knowledge point matrix and the user-test question matrix are input into a cognitive diagnosis model, so that the cognitive diagnosis model can obtain the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point; the user's weak knowledge points are knowledge points whose comprehensive mastery is less than a preset threshold; Scan the knowledge point set using the AprioriAll algorithm, combine the user's weak knowledge points, and generate a recommended learning path by increasing the weight value of the user's weak knowledge points during the learning path planning process; A plurality of member search engines are introduced to perform resource retrieval on the recommended learning path, and the retrieval results of each member search engine are integrated to obtain a recommendation result.

2. The educational resource retrieval method according to claim 1, characterized in that: The cognitive diagnosis model includes a VFuzzy-CDF model, a CFuzzy-CDF model and a Fuzzy-CDF model. The test question-knowledge point matrix and the user-test question matrix are input into the cognitive diagnosis model, so that the cognitive diagnosis model can obtain the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point, including: Inputting the video-knowledge point matrix and the user-video matrix into a VFuzzy-CDF model, so that the VFuzzy-CDF model obtains the user's first weak knowledge point by judging the comprehensive mastery of each knowledge point by each user; Input the other resource-knowledge point matrix and the user-other resource matrix into the CFuzzy-CDF model, so that the CFuzzy-CDF model can obtain the second weakest knowledge point of each user by judging the comprehensive mastery of each knowledge point by each user; Inputting the exercise-knowledge point matrix and the user-exercise matrix into a fuzzy-CDF model, so that the fuzzy-CDF model can obtain the third weakest knowledge point of each user by judging the comprehensive mastery of each knowledge point by each user; The user's first weak knowledge point, the user's second weak knowledge point and the user's third weak knowledge point are integrated to obtain the user's weak knowledge point.

3. The educational resource retrieval method according to claim 1, characterized in that: The test question-knowledge point matrix and the user-test question matrix are input into a cognitive diagnosis model, so that the cognitive diagnosis model obtains the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point in the following manner: Obtaining the fuzzy attribute weight of each test question to each knowledge point according to the test question-knowledge point matrix; The fuzzy attribute weight is used to indicate the relevance of each test question to each knowledge point; According to the fuzzy attribute weight of each test question for each knowledge point and each user's mastery of each knowledge point, the mastery degree of each user on the required knowledge point in each test question is obtained; Based on each user's mastery of the knowledge points required in each test question, the user's error rate and guessing rate are introduced to obtain each user's actual score rate; The comprehensive mastery of each knowledge point by the user is determined by combining the actual score rate of each user and the probability of the user's correct answer to each question in the question-knowledge point matrix, and the knowledge points with a comprehensive mastery level less than a preset threshold are regarded as the user's weak knowledge points.

4. The educational resource retrieval method according to claim 1, characterized in that: Introduce multiple member search engines to search for resources on the recommended learning path, and integrate the search results of each member search engine to obtain recommendation results, including: Introducing multiple member search engines to perform resource retrieval on the recommended learning path, and integrating the retrieval results of each member search engine to obtain an initial retrieval result; After removing invalid search results from the initial search results, the remaining search results are sorted to obtain recommended results.

5. The educational resource retrieval method according to claim 4, characterized in that: After removing invalid search results from the initial search results, the remaining search results are sorted to obtain recommended results, including: After removing invalid search results from the initial search results, the remaining search results are sorted in combination with the accuracy, recall and nDCG of each member search engine to obtain recommendation results.

6. An educational resource retrieval device based on cognitive diagnosis and learning path, characterized in that: The educational resource retrieval device comprises: A matrix determination module is used to determine a test question-knowledge point matrix and a user-test question matrix according to a knowledge point set; the test question-knowledge point matrix includes a video-knowledge point matrix, an exercise-knowledge point matrix and an other resource-knowledge point matrix; the user-test question matrix includes a user-video matrix, a user-exercise matrix and a user-other resource matrix; the test question-knowledge point matrix is ​​used to indicate the degree of association between the test question and the knowledge point; the user-test question matrix is ​​used to indicate the user's answer to the test question; A user weak knowledge point determination module is used to input the test question-knowledge point matrix and the user-test question matrix into a cognitive diagnosis model, so that the cognitive diagnosis model can obtain the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point; the user's weak knowledge points are knowledge points whose comprehensive mastery is less than a preset threshold; A recommended learning path generation module, used to scan the knowledge point set using the AprioriAll algorithm, and generate a recommended learning path by increasing the weight value of the user's weak knowledge points in the learning path planning process in combination with the user's weak knowledge points; The search result integration module is used to introduce multiple member search engines to perform resource search on the recommended learning path, and integrate the search results of each member search engine to obtain a recommendation result.

7. The educational resource retrieval device for cognitive diagnosis and learning path according to claim 6, characterized in that: The cognitive diagnosis model includes a VFuzzy-CDF model, a CFuzzy-CDF model and a Fuzzy-CDF model. The user weak knowledge point determination module is specifically used to input the video-knowledge point matrix and the user-video matrix into the VFuzzy-CDF model, so that the VFuzzy-CDF model obtains the user's first weak knowledge point by judging the comprehensive mastery of each user on each knowledge point; input the other resource-knowledge point matrix and the user-other resource matrix into the CFuzzy-CDF model, so that the CFuzzy-CDF model obtains the user's second weak knowledge point by judging the comprehensive mastery of each user on each knowledge point; input the exercise-knowledge point matrix and the user-exercise matrix into the Fuzzy-CDF model, so that the Fuzzy-CDF model obtains the user's third weak knowledge point by judging the comprehensive mastery of each user on each knowledge point; and integrate the user's first weak knowledge point, the user's second weak knowledge point and the user's third weak knowledge point to obtain the user's weak knowledge point.

8. The educational resource search device according to claim 6, characterized in that: The user weak knowledge point determination module inputs the test question-knowledge point matrix and the user-test question matrix into a cognitive diagnosis model, so that the cognitive diagnosis model obtains the user's weak knowledge points by judging the comprehensive mastery of each user on each knowledge point in the following manner: Obtaining the fuzzy attribute weight of each test question to each knowledge point according to the test question-knowledge point matrix; The fuzzy attribute weight is used to indicate the relevance of each test question to each knowledge point; According to the fuzzy attribute weight of each test question for each knowledge point and each user's mastery of each knowledge point, the mastery degree of each user on the required knowledge point in each test question is obtained; Based on each user's mastery of the knowledge points required in each test question, the user's error rate and guessing rate are introduced to obtain each user's actual score rate; The comprehensive mastery of each knowledge point by the user is determined by combining the actual score rate of each user and the probability of the user's correct answer to each question in the user-question matrix, and the knowledge points with a comprehensive mastery level less than a preset threshold are regarded as the user's weak knowledge points.

9. The educational resource search device according to claim 6, characterized in that: The retrieval result integration module is specifically used to introduce multiple member search engines to perform resource retrieval on the recommended learning path, and integrate the retrieval results under each member search engine to obtain the initial retrieval result; after removing invalid retrieval results from the initial retrieval result, the remaining retrieval results are sorted to obtain the recommended result.

10. The educational resource search device according to claim 9, characterized in that: The search result integration module removes invalid search results from the initial search results and sorts the remaining search results to obtain recommended results, including: After removing invalid search results from the initial search results, the remaining search results are sorted in combination with the accuracy, recall and nDCG of each member search engine to obtain recommendation results.

Images