A knowledge point sorting method based on multi-dimensional information

Abstract

The present application relates to the technical field of teacher teaching, especially to a knowledge point sorting method based on multi-dimensional information. The steps are as follows: constructing initial knowledge point information through knowledge point acquisition; constructing a statistical basic graph through the frequency of knowledge points in the initial knowledge point information; sorting according to the textbook directory hierarchical structure and statistical information weighting; and constructing a machine learning classification algorithm model based on simple machine learning logistic regression. The present application provides a knowledge point sorting method based on multi-dimensional information, which comprehensively considers the importance of knowledge points based on multiple dimensions of data through statistical information arrangement, and at the same time constructs a classification model to classify the importance of knowledge points in business; the PageRank algorithm is used for accurate scoring, which ensures that the knowledge point directory structure is quickly interacted with the subject professional personnel and the model is continuously iterated, improves the labeling efficiency, and at the same time ensures the accuracy of the knowledge point sorting.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of teacher teaching, and in particular to a knowledge point sorting method based on multi-dimensional information. BACKGROUND

[0002] The existing knowledge point sorting system is mostly based on pre-training of a large amount of data, then obtaining a feature vector of the knowledge point, and using the feature vector to analyze the correlation between different knowledge points;

[0003] For example, Chinese invention patent (CN110599377A) discloses a knowledge point sorting method and device for online learning, which solves the technical problem that the knowledge points of an online programming platform are based on a knowledge network or a knowledge graph, but the knowledge network and the knowledge graph cannot express the dispersion and implicitness of the knowledge points, resulting in inaccurate personalized recommendation of the knowledge points;

[0004] For example, Chinese invention patent (CN114493944A) discloses a knowledge point sorting method and device for online learning, which solves the problem that the current method cannot accurately diagnose the weak knowledge points of students and cannot further plan a reasonable learning path according to the individual situation of students, thereby causing low learning efficiency;

[0005] At the same time, patents CN110599377A and CN114493944A both put the subject knowledge points into an NLP large model for diagnosis, construct a multi-layer relationship network according to the feature vectors of users or different subject knowledge points, and determine the importance degree of the knowledge points. This technology is relatively simple and relies on a large amount of data, but the sorting obtained is either based on the importance score of the pre-training model or depends on the relatively broad knowledge point sorting obtained by the collaborative relationship network, and lacks actual business scenarios to judge the score order of whether the knowledge points are core knowledge points or ordinary knowledge points;

[0006] For example, Chinese invention patent (CN202211325473) discloses a full-text knowledge point extraction method based on an electronic book, which solves the technical problem that the existing technology cannot determine the boundary of the key word and accurately extract the knowledge points, and the effect is not good in the actual scene;

[0007] Therefore, we design a knowledge point sorting method based on multi-dimensional information to provide another technical solution to the above technical problems. SUMMARY

[0008] Therefore, it is necessary to provide a knowledge point sorting method based on multi-dimensional information to solve the technical problems proposed in the background art.

[0009] In order to solve the above technical problems, the present application adopts the following technical solutions:

[0010] A knowledge point sorting method based on multi-dimensional information, the steps are as follows:

[0011] An initial knowledge point information is constructed by acquiring knowledge points;

[0012] A statistical basic graph is constructed by the frequency of the occurrence of knowledge points in the initial knowledge point information;

[0013] According to the textbook directory hierarchical structure and statistical information weighting, sorting is performed;

[0014] A machine learning classification algorithm model based on simple machine learning logistic regression is constructed;

[0015] A PageRank sorting algorithm is constructed based on the directory hierarchy of textbooks and encyclopedias.

[0016] As a preferred embodiment of the knowledge point sorting method based on multi-dimensional information provided by the application, the knowledge points are acquired by textbook fragments, encyclopedia definitions, Zhihu documents and Zhiwen documents to obtain initial knowledge point information.

[0017] As a preferred embodiment of the knowledge point sorting method based on multi-dimensional information provided by the application, the statistical basic graph is constructed by co-occurrence relationship module, species relationship module and correlation relationship module to count the frequency of the occurrence of knowledge points;

[0018] The co-occurrence relationship module records the same textbook fragments, encyclopedia information and Zhiwen information that appear simultaneously and is arranged in descending order according to frequency;

[0019] The species relationship module records the occurrence frequency of the upper and lower structure of the directory;

[0020] The correlation relationship module is used to record the information frequency of the unrelated textbooks and the related information in semantics, word entries, etc.

[0021] As a preferred embodiment of the knowledge point sorting method based on multi-dimensional information provided by the application, the sorting according to the textbook directory hierarchical structure and statistical information weighting obtains initial core knowledge points, associated knowledge points and attribute knowledge points;

[0022] The core knowledge points are used to count the main level of the directory, the encyclopedia entries and the Zhiwen document information;

[0023] The associated knowledge points are used to count the lower structure of the directory, the Zhiwen content information and the subheadings;

[0024] The attribute knowledge points are used to count the low frequency of occurrence, the existence of the encyclopedia entries and the multiple Zhiwen information.

[0025] As a preferred embodiment of the knowledge point sorting method based on multi-dimensional information provided by the application, the machine learning classification algorithm model is constructed according to initial knowledge point information and statistical frequency information in a statistical basic graph.

[0026] As a preferred embodiment of the knowledge point sorting method based on multi-dimensional information provided by the application, the machine learning classification algorithm model adopts a logistic regression, and a logistic regression expression is as follows:

[0027]

[0028]

[0029] wherein X is a random variable, μ is a mean of a variable distribution, γ is a variance of the variable distribution, and a distribution function represents a probability (0, 1) that the variable X is less than x;

[0030] According to the knowledge point information of the selected subject, the logistic regression is used to obtain probabilities that the knowledge points belong to core knowledge points, attribute knowledge points and irrelevant knowledge points, and the knowledge point order is rearranged.

[0031] As a preferred embodiment of the knowledge point sorting method based on multi-dimensional information provided by the application, a PageRank sorting algorithm is constructed, and the steps are as follows:

[0032] The core knowledge points and the attribute knowledge points are combined with a baike level and a textbook level.

[0033] Each directory of the textbook corresponds to a page of knowledge points, and each knowledge point directory structure of the baike corresponds to a page.

[0034] As a preferred embodiment of the knowledge point sorting method based on multi-dimensional information provided by the application, a PageRank algorithm is used to construct an importance sorting of the knowledge points relative to the directory structure, and the steps are as follows:

[0035] S1. An initial PageRank value is assigned to the knowledge points, and PageRank values of all the knowledge points are initialized to the same value, 0.8.

[0036] S2. The PageRank value of each knowledge point is calculated by traversing the entire network graph.

[0037] S3. A calculation formula of the PageRank value is as follows:

[0038] PR(i) = (1-d) + d*(PR(j) / L(j))

[0039] Wherein, d is a damping factor (usually set to 0.85), L(j) refers to the number of directories pointing to knowledge point j;

[0040] S4. Repeat the execution of step S2 and step S3 until convergence determines when to stop iteration;

[0041] S5. The final PageRank value of each knowledge point is ranked according to the final PageRank value, and the ranking result is returned to the professional human-computer interaction, and the correctness of the ranking result is determined through human-computer interaction.

[0042] As a preferred embodiment of the knowledge point sorting method based on multi-dimensional information provided by the application, in the S4 step, the execution of step S2 and step S3 is repeated until the convergence threshold 0.001 is reached to determine when to stop iteration.

[0043] As a preferred embodiment of the knowledge point sorting method based on multi-dimensional information provided by the application, in the S5 step, if the judgment is wrong, the ranking result is corrected, and the corrected ranking result is sent to the machine learning classification algorithm model for iterative correction;

[0044] If the result is correct, the ranking result is used.

[0045] It can be seen without doubt that the above technical solutions of the application can certainly solve the technical problems to be solved by the application.

[0046] At the same time, through the above technical solutions, the application at least has the following beneficial effects:

[0047] The knowledge point sorting method based on multi-dimensional information provided by the application sorts the knowledge points based on the importance of the knowledge points comprehensively considered based on the data of multiple dimensions through the statistical information arrangement, constructs the co-occurrence, correlation and species relationship based on the information of textbooks, encyclopedias and the like, avoids the way of obtaining wide sorting by relying on a large model, constructs a classification model at the same time, classifies the importance of the knowledge points in business, and finally scores accurately by using the PageRank algorithm according to the directory level of the textbooks and the directory structure of the encyclopedias, guarantees the knowledge point directory structure, and quickly interacts with the subject professional personnel and iterates the model continuously, improves the labeling efficiency, and guarantees the accuracy of the knowledge point sorting. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0049] Fig. 1 Statistical information diagram of the present application;

[0050] Fig. 2 Network structure diagram of the present application;

[0051] Fig. 3 Flow diagram of the present application. DETAILED DESCRIPTION

[0052] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0053] In order to make the persons skilled in the art better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0054] It should be noted that the embodiments in the present application and the features and technical solutions in the embodiments can be combined with each other without conflict.

[0055] It should be noted that: similar reference numerals and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0056] Referring to Figs. 1-3 A knowledge point sorting method based on multi-dimensional information, the steps are as follows:

[0057] Constructing initial knowledge point information, obtaining knowledge points through textbook fragments, encyclopedia definitions, Zhihu documents, and Zhiwen documents.

[0058] For example, the main knowledge point network layer, the associated knowledge point network layer function, the network layer service, the network layer device, the attribute knowledge point data report service, the virtual circuit service, the network layer service characteristics, and the statistical basic information diagram of the associated information of each knowledge point;

[0059] The frequency of the knowledge points is constructed by the co-occurrence relationship module, the species relationship module, and the correlation relationship module to construct a statistical basic diagram;

[0060] The co-occurrence relationship module records the same textbook fragments, encyclopedia information, Zhiwen information, etc. that appear simultaneously and are arranged in descending order according to frequency;

[0061] The species relationship module records the occurrence frequency of the directory's superior-inferior structure, and the knowledge points in the directory main structure and content are in species relationship;

[0062] The correlation module is used for recording the information frequency of the correlation in semantics, entries, etc., such as the lack of correlation of the teaching material, the appearance of the same entry of the encyclopedic information or the simultaneous appearance of the MOOC information;

[0063] Thus, all knowledge points are obtained according to the above relationship, based on the teaching material directory and multi-dimensional information, to obtain the initial statistical basic information graph;

[0064] Then, according to the teaching material hierarchical structure, the first-level directory, the second-level directory, the third-level directory, the content, and the statistical information weighting, the initial core knowledge points (the directory main level, the encyclopedic entry and the information of the HowNet literature), the associated knowledge points (the directory lower structure, the HowNet content information and the subheading), and the attribute knowledge points (the low appearance frequency, but the encyclopedic entry exists and the HowNet information is much) are obtained;

[0065] Then, according to the above initial information and the statistical frequency information, a machine learning classification algorithm model based on a simple machine learning logistic regression is constructed, and the input features are:

[0066] Knowledge point hierarchical structure (teaching material hierarchy), knowledge point appearance teaching material segment frequency, encyclopedic information number, HowNet resource quantity, internal MOOC resource quantity, co-occurrence knowledge point quantity, subset knowledge point quantity, related knowledge point quantity

[0067] Classification: core knowledge point, attribute knowledge point, irrelevant knowledge point

[0068] The present application adopts a lightweight model: logistic regression, and a logistic regression formula:

[0069] Distribution function:

[0070]

[0071] Density function:

[0072]

[0073] Wherein: X is a random variable, μ is the mean of the variable distribution, and γ is the variance of the variable distribution; the distribution function represents the probability (0, 1) that the variable X is less than x.

[0074] According to the knowledge point information of the selected subject, the probability that the knowledge point belongs to the core knowledge point, the attribute knowledge point and the irrelevant knowledge point is obtained by using the logistic regression, and the order of the knowledge points is rearranged;

[0075] Finally, in the core knowledge points, attribute knowledge points combined with encyclopedia level and textbook level, that is, each directory of the textbook corresponds to a page of knowledge points, and each knowledge point directory structure of the encyclopedia corresponds to a page. The importance of knowledge points relative to the directory structure is sorted by using the PageRank algorithm. The process is as follows:

[0076] 1. First, assign an initial PageRank value to the knowledge points, and initialize the PageRank values of all knowledge points to the same value, which is 0.8.

[0077] 2. Then, the PageRank value of each knowledge point is calculated by traversing the entire network graph. In this process, for each knowledge point i, the contribution of all other knowledge points j pointing to it to its PageRank value needs to be calculated.

[0078] 3. The calculation of the PageRank value is based on the following formula: PR(i) = (1-d) + d*(PR(j) / L(j)), where d is a damping factor (usually set to 0.85), and L(j) is the number of directories pointing to knowledge point j.

[0079] 4. Repeat steps 2 and 3 until convergence, or set a convergence threshold of 0.001 to determine when to stop iteration.

[0080] 5. Finally, according to the final PageRank value of each knowledge point, they are sorted and returned to our professional human-computer interaction as the sorting result. Through human-computer interaction, the sorting result is judged whether it is correct or not. If it is wrong, the sorting result is corrected, and the corrected sorting result is fed into the machine learning classification algorithm model for iterative correction. If the result is correct, the sorting result is used. Through rapid interaction with subject professionals and continuous iteration of the model, the efficiency of labeling is improved while ensuring the accuracy of knowledge point sorting, realizing the optimal order of interaction with labeling personnel, realizing the importance of multi-dimensional information of knowledge points in teaching, and having initial knowledge point relationships, which has important significance in the field of teaching.

[0081] The effect is as follows:

[0082] 1. Utilize NLP related technologies including pre-trained model feature vectors, roughly according to textbook fragments, encyclopedia information, know network, blog and other network information, build knowledge point information network (network graph), for example: knowledge points and knowledge point aliases, knowledge points and knowledge point co-occurrence (same fragment), knowledge points and knowledge points in the same encyclopedia, etc.; Then use machine learning algorithm to classify and score each node in the information network to determine the initial score of the importance of the knowledge point; Finally, according to the importance of the node and the directory level relationship of the teaching material, use PageRank algorithm, according to different disciplines, accurate sorting, and according to the interaction with the labeling personnel and the customized setting of the discipline, build a multi-dimensional information knowledge point importance sorting system in the education and teaching scene.

[0083] 2. Adopt multi-dimensional data of disciplines, textbook fragments, encyclopedia information, and other information, build knowledge point information network according to statistical methods, build initial classification data, core knowledge points, important knowledge points, and attribute knowledge points; Build a classification model to classify the importance of knowledge points, and according to the hierarchical directory structure of the teaching material, use PageRank algorithm for accurate scoring; Finally, interact with the recommended subject knowledge point sorting results and professional subject personnel to build an accurate multi-dimensional information knowledge point sorting system in the teaching scene.

[0084] 3. It can consider the importance of knowledge points based on multiple dimensions of statistical information arrangement, build co-occurrence, correlation, and species relationship based on teaching materials, encyclopedia, etc. to avoid relying heavily on large models to obtain broad sorting; At the same time, build a classification model to classify the importance of knowledge points; Finally, according to the directory level of the teaching material and the directory structure of the encyclopedia, use PageRank algorithm for accurate scoring to ensure that the knowledge point directory structure is quickly interacted with the subject professional personnel and the model is continuously iterated to improve the annotation efficiency while ensuring the accuracy of the knowledge point sorting.

[0085] 4. In the field of teaching, integrate dimension information for specified disciplines, combine basic statistical information for rough sorting, build core knowledge points and attribute knowledge points; Then use logistic regression to get the sorting score of each knowledge point in real time, use the hierarchical structure of the knowledge point itself to build PageRank sorting rules to get the final knowledge point importance sorting, which realizes the optimal order and interaction with the labeling personnel in the case of ensuring the hierarchical structure of the knowledge point, realizes the importance sorting of multi-dimensional information of knowledge points in teaching, and has initial knowledge point relationship, which is of great significance in the field of teaching.

[0086] The preferred embodiments of the application disclosed above are only to facilitate the elucidation of the application. The preferred embodiments do not describe all the details of the application, nor limit the application to the specific embodiments described. Obviously, many modifications and variations are possible in light of the above teachings. The embodiments were chosen and described in order to provide the best illustration of the principles of the application and its practical application, and to thereby enable one of ordinary skill in the art to best utilize the application. The application is only to be limited by the claims and their full scope and equivalents.

Claims

1. A method for sorting knowledge points based on multi-dimensional information, characterized in that, The steps are as follows: Construct initial knowledge point information through knowledge point acquisition; Construct statistical basic graph through the frequency of knowledge points in initial knowledge point information; Sort according to the textbook directory hierarchical structure and statistical information weighting; Build a machine learning classification algorithm model based on simple machine learning logistic regression; Build a PageRank sorting algorithm based on the directory hierarchy of textbooks and encyclopedias; Sort according to the textbook directory hierarchical structure and statistical information weighting to obtain initial core knowledge points, associated knowledge points, and attribute knowledge points. The core knowledge points are used to count the main level of the directory, the encyclopedia entry, and the literature information of the HowNet; The associated knowledge points are used to count the lower level structure of the directory, the HowNet content information, and the subheadings; The attribute knowledge points are used to count the low frequency of occurrence, the existence of the encyclopedia entry, and the multiple HowNet information. The feature is to build a PageRank sorting algorithm, and the steps are as follows: Combine the core knowledge points, attribute knowledge points, encyclopedia hierarchy, and textbook hierarchy; Make each directory of the textbook correspond to a knowledge point page, and each knowledge point directory structure of the encyclopedia correspond to a page; Use the PageRank algorithm to build the importance sorting of knowledge points relative to the directory structure, and the steps are as follows: S1. Assign an initial PageRank value to the knowledge point, and initialize the PageRank values of all knowledge points to the same value, which is 0.8; S2. Calculate the PageRank value of each knowledge point by traversing the entire network graph; S3. The calculation formula of the PageRank value is as follows: PR(i)=(1-d)+d*(PR(j) / L(j)) Where d is a damping factor (usually set to 0.85), and L(j) is the number of directories pointing to knowledge point j; S4. Repeat steps S2 and S3 until convergence to determine when to stop iteration; S5. According to the final PageRank value of each knowledge point, sort them and return them as sorting results to our professional human-computer interaction, and judge whether the sorting results are correct through human-computer interaction. 2.The knowledge point sorting method based on multi-dimension information according to claim 1, characterized in that, The knowledge point acquisition obtains initial knowledge point information through textbook fragments, encyclopedia definitions, Zhihu documents, and HowNet documents. 3.The knowledge point sorting method based on multi-dimension information according to claim 1, characterized in that, The construction of statistical basic graph is based on the frequency statistics of knowledge points through co-occurrence relationship module, species relationship module, and correlation relationship module; The co-occurrence relationship module records the same textbook fragments, encyclopedia information, and HowNet information that appear simultaneously in descending order of frequency; The species relationship module records the occurrence frequency of the directory's upper and lower structures; The correlation relationship module is used to record the information frequency of the unrelated textbooks and the related information in semantics, entries, etc. 4.The method of claim 1, wherein, The construction of the machine learning classification algorithm model is based on the initial knowledge point information and the statistical frequency information in the statistical basic graph.

5. The method of claim 1, wherein, The machine learning classification algorithm model uses logistic regression, and the logistic regression expression is as follows: wherein, is a random variable, is a mean of the distribution of the variable, is a variance of the distribution of the variable; the distribution function denotes a probability (0, 1) of the variable . Using logistic regression, according to the knowledge point information of the selected subject, the probability of the knowledge point belonging to core knowledge points, attribute knowledge points, and irrelevant knowledge points is obtained, and the order of the knowledge points is rearranged.

6. The knowledge point sorting method based on multi-dimension information according to claim 1, characterized in that, In the S4 step, steps S2 and S3 are repeatedly executed until a convergence threshold value 0.001 is reached to determine when to stop iteration.

7. The method of claim 1, wherein, In the S5 step, if the result is incorrect, the ranking result is corrected, and the corrected ranking result is fed to the machine learning classification algorithm model for iterative correction. If the result is correct, the ranking result is used.

Images