CN117557425B

CN117557425B - Question bank data optimization method and system based on intelligent question bank system

Info

Publication number: CN117557425B
Application number: CN202311675199.3A
Authority: CN
Inventors: 黎国权; 朱晖
Original assignee: Guangzhou Xiaoma Zhixue Technology Co ltd
Current assignee: Guangzhou Xiaoma Zhixue Technology Co ltd
Priority date: 2023-12-08
Filing date: 2023-12-08
Publication date: 2024-04-16
Anticipated expiration: 2043-12-08
Also published as: CN117557425A

Abstract

The embodiment of the application provides a question bank data optimization method and a system based on an intelligent question bank system, which effectively determine and optimize butt joint nodes among elements in a question bank by acquiring and processing embedded representation data of a target question bank entity and a knowledge point entity and associated network information. Specifically, first, the test points of the stem and knowledge point entities of the target question bank entity are utilized to construct a first association network and a second association network, and then the question bank space logic point of the target docking node is established based on embedded representation data and entity inter-association information. Thereafter, two updating operations are performed: generating a preliminary intelligent question bank frame map based on the target docking node by the first updating; and the logic points are further refined after the second updating, so that a second intelligent question bank frame map with higher precision is generated. Through the multi-stage optimization, the positioning accuracy and the recommended quality of the learning resources can be remarkably improved.

Description

Question bank data optimization method and system based on intelligent question bank system

Technical Field

The application relates to the technical field of computer information, in particular to a question bank data optimization method and system based on an intelligent question bank system.

Background

In the technical field of education, the intelligent question bank system is used as a core resource for offline learning and teaching, and plays a vital role in improving the education quality and efficiency. However, conventional question bank systems suffer from a number of limitations, the most significant of which is the inability to accurately capture the logical relationships between questions and knowledge points, resulting in a lack of personalization and contextual relevance of learning path recommendations. In addition, these systems often employ static content indexing and simple keyword matching methods, which are difficult to dynamically adapt to changing educational needs and learning content.

Due to the lack of highly-refined data processing and intelligent map generation mechanisms, the conventional question bank system is difficult to effectively integrate a large number of scattered learning resources, and the question bank structure cannot be optimized by fully utilizing data mining and artificial intelligence technologies. This results in the formation of islands of topic library information, making it difficult for learners to discover and utilize interdisciplinary or interdisciplinary knowledge connections, thereby reducing learning efficiency.

Disclosure of Invention

In order to overcome the above-mentioned shortcomings in the prior art at least, the present application aims to provide a method and a system for optimizing question bank data based on an intelligent question bank system, which can more accurately reflect and utilize complex correlations between questions and related knowledge points thereof, and simultaneously provide a flexible and personalized learning resource organization mode so as to support more efficient and targeted teaching and learning activities. Through the deep analysis and intelligent management of the logic relationship between the question bank entities, the configuration efficiency and the use effect of the education resources can be remarkably improved, and the requirements of the modern education technology on individuation and intelligence are met.

In a first aspect, the present application provides a method for optimizing question bank data based on an intelligent question bank system, which is applied to a learning platform system, and the method includes:

Acquiring first embedded representation data of each target question bank entity in an intelligent question bank system, and acquiring second embedded representation data of each knowledge point entity, wherein the first representation element in the target question bank entity is a question stem, and the second representation element in the knowledge point entity is a test point;

extracting a first association network of the target question bank entity and a second association network of the knowledge point entity, wherein a connection line between the first expression elements in the first association network is a logic relationship between the questions;

determining a question bank space logic point of a target butt joint node of the target question bank entity and the knowledge point entity according to the first embedded representation data, the second embedded representation data and entity mutual association information;

according to the question bank space logic points of each target docking node, first updating is carried out on the question bank space logic points of each element in the first association network and the second association network, and a first intelligent question bank frame map is generated;

And carrying out second updating on the spatial logic points of the question bank of each element in the first intelligent question bank frame map according to the element interrelation information of each element in the first intelligent question bank frame map, and generating a second intelligent question bank frame map, wherein the precision of the second updating is larger than that of the first updating.

In a second aspect, an embodiment of the present application further provides a learning platform system, where the learning platform system includes a processor and a machine-readable storage medium, where the machine-readable storage medium stores a computer program, where the computer program is loaded and executed according to the processor to implement the method for optimizing question bank data based on the intelligent question bank system in the first aspect.

According to the technical scheme of any aspect, the butt joint nodes among the elements in the question bank are effectively determined and optimized by acquiring and processing the embedded representation data of the target question bank entity and the knowledge point entity and the associated network information. Specifically, first, the test points of the stem and knowledge point entities of the target question bank entity are utilized to construct a first association network and a second association network, and then the question bank space logic point of the target docking node is established based on embedded representation data and entity inter-association information. Thereafter, two updating operations are performed: generating a preliminary intelligent question bank frame map based on the target docking node by the first updating; and the logic points are further refined after the second updating, so that a second intelligent question bank frame map with higher precision is generated. Through the multi-stage optimization, the positioning accuracy and the recommendation quality of learning resources can be remarkably improved, so that personalized learning is promoted and the education effect is improved.

Specifically, the application effectively constructs the association network comprising the question stems (first representing elements) and the examination points (second representing elements) by acquiring the first embedded representing data of the target question bank entity and the second embedded representing data of each knowledge point entity, and further extracts the first association network and the second association network between the target question bank entity and the knowledge point entity, wherein the first association network reflects the logical relationship between the question stems, so that the relationship between the knowledge points is more clear and visible.

By comprehensively applying the first embedded representation data and the second embedded representation data and combining entity interrelation information, the target butt joint node between the target question bank entity and the knowledge point entity is accurately determined, so that the optimal logic point of each element is established in the question bank space. The process not only enhances the tightness of the butt joint between the question bank entities, but also provides a solid foundation for the subsequent study path planning.

And according to the question bank space logic points of the target docking node, performing first updating on the first association network and the second association network for the first time, and generating a preliminary first intelligent question bank frame map. Then, a second update is performed using the cross-correlation information of the elements in the frame map, resulting in a second intelligent question bank frame map of higher accuracy. The secondary updating flow obviously improves the accuracy and the reliability of the system, so that the question bank has more adaptability and individuation characteristics, and the configuration and the utilization efficiency of educational resources are greatly optimized.

In a word, the application obviously improves the logic point positioning precision of the intelligent question bank through a two-stage updating process, thereby providing more accurate learning advice for students, providing more effective teaching auxiliary tools for teachers and promoting the intelligent management of educational resources.

Drawings

For a clearer description of the technical solutions of the embodiments of the present application, reference will be made to the accompanying drawings, which are needed to be activated in the embodiments, it being understood that the following drawings only illustrate some embodiments of the present application and are therefore not to be considered limiting of the scope, and that other related drawings can be obtained according to these drawings without the inventive effort for a person skilled in the art.

FIG. 1 is a flow chart of a method for optimizing question bank data based on an intelligent question bank system according to an embodiment of the application;

Fig. 2 is a schematic functional block diagram of a learning platform system for implementing the method for optimizing question bank data based on the intelligent question bank system according to an embodiment of the present application.

Detailed Description

The following description is presented to enable one of ordinary skill in the art to make and use the application and is provided in the context of a particular application and its requirements. It will be apparent to those having ordinary skill in the art that various changes can be made to the disclosed embodiments and that the general principles defined herein may be applied to other embodiments and applications without departing from the principles and scope of the application. Therefore, the present application is not limited to the described embodiments, but is to be accorded the widest scope consistent with the claims.

Referring to fig. 1, the application provides a method for optimizing question bank data based on an intelligent question bank system, which comprises the following steps.

Step S110, first embedded representation data of each target question bank entity in the intelligent question bank system are obtained, second embedded representation data of each knowledge point entity are obtained, the first representation element in the target question bank entity is a question stem, and the second representation element in the knowledge point entity is a test point.

For example, this step includes digitizing stems of all questions within a particular question bank (the target question bank entity) to obtain their embedded representations (i.e., first embedded representation data). At the same time, all knowledge points involved need to be processed identically to get their embedded representation (i.e. the second embedded representation data).

Illustratively, the target question bank entity refers to a single question bank, each containing a plurality of questions. The first embedded representation data is the embedded representation data obtained by digitizing the stem of each question in the question bank by the pointer, and is usually a vector by converting the text corresponding to the object question bank entity into a numerical form which can be understood and processed by the computer. The knowledge point entity refers to specific knowledge points defined in the education or examination outline, and each knowledge point corresponds to specific learning content. The second embedded representation data refers to the embedded representation data obtained by digitizing the description of each knowledge point, and is also a vector capable of representing the content of the knowledge point.

For each target question bank entity (i.e., each question bank), the stem of all questions therein may be analyzed. For example, a question bank may be focused on junior mathematics, and there are various problems related to mathematics, each with its own stem, such as "solve the unitary once equation ax+b=0".

For each topic stem, natural language processing techniques may be used to generate its first embedded representation data. This process involves text analysis, feature extraction, etc., ultimately producing a vector that can represent the meaning of each stem. For example, an embedded representation of the stem solving the unitary once equation ax+b=0 would capture features related to algebraic equations.

At the same time, it is also necessary to obtain a description of each knowledge point entity and generate second embedded representation data for it. Assuming that a knowledge point is a "solution to the unitary one-time equation," an embedded representation is generated for the descriptive text of the knowledge point. Through these two embedded representations, a relationship between the topic and the knowledge point can be established in the computer. For example, if the stem embedded representation of a topic is very close to the embedded representation of a knowledge point, this means that the topic may be used to examine the knowledge point.

The first embedded representation data and the second embedded representation data allow for conversion of text data into numeric data that can be processed by an algorithm throughout the process. These embedded representation data exist in the form of vectors in a multidimensional space where similar stems or knowledge points are located closer together in space. By analyzing the distance and direction between these vectors, the relationship between the questions and knowledge points can be found, thereby organizing and optimizing the question bank to more effectively serve educational and assessment purposes.

Step S120, extracting a first association network of the target question bank entity and a second association network of the knowledge point entity, where a connection line between the first expression elements in the first association network is a logical relationship between the questions stems.

For example, in this step, two types of association networks may be constructed: a correlation network (first correlation network) of topics in the topic library, wherein the first expression element represents a single topic, and the connection line represents a logical relationship between topics, such as similarity, progression of difficulty level and the like; and a correlation network (second correlation network) of knowledge points, wherein the second representation element represents a single knowledge point and the connection lines represent logical or conceptual connections between knowledge points.

Illustratively, the first association network may also be referred to as a topic association network of target topic library entities), and for a particular target topic library entity (i.e., a particular topic library), the first association network is made up of all topics in the topic library. In this first association network, each first presentation element represents a separate topic, and the links between the first presentation elements represent logical relationships between topics.

For example, assume that there is a junior middle school math question bank. The question bank contains different questions such as:

-title a: "solve the unitary once equation ax+b=0. "

-Title B: "if the two straight lines are parallel, it is verified that their slopes are equal. "

-Title C: "solve the following ratio problem: if the total weight of 3 apples and 2 oranges is 1 kg, the total weight of 5 apples and 4 oranges is calculated. "

When the first association network is constructed, the problem stems of the problems can be analyzed, and whether connection lines exist between nodes or not and the weight of the connection lines are determined according to the similarity between the problems, the difficulty, the related knowledge points and other factors. For example, both topic A and topic C may involve algebraic knowledge, and thus there may be a link between them in the network. The topic B involves geometric knowledge, so that its connection to topics a and C may be weak or non-existent.

The second association network may also be referred to as a knowledge point association network, which is made up of all knowledge points defined. In this second association network, each second representing element represents a separate knowledge point, and the links between the second representing elements represent logical relationships between the knowledge points.

Continuing with the example above, there may be several points of knowledge:

knowledge point X: "solution to the unitary once-through equation". "

Knowledge point Y: "nature of straight line and slope concept". "

Knowledge point Z: "solution to the proportional problem". "

When the second association network is constructed, the connection line can be determined according to the relation between the knowledge points. For example, knowledge point X and knowledge point Z both fall within algebraic categories, so there may be strong links between them in the network. Knowledge point Y belongs to the geometric category and the line connecting the first two knowledge points may be relatively weak.

By constructing the two association networks, the relation between topics and knowledge points in the topic library can be better understood. The relation can help to perform tasks such as question recommendation, knowledge point reinforcement, teaching path planning and the like. At the same time, the first and second association networks can also be used as a basis for subsequent understanding of the topic library structure for further data analysis and machine learning applications.

Step S130, determining a question bank space logical point of the target butt joint node of the target question bank entity and the knowledge point entity according to each first embedded representation data, each second embedded representation data and the entity correlation information.

For example, this step aims at determining the relationship between topics in the target topic library entity and the knowledge point entity. By analyzing each of the first embedded representation data and each of the second embedded representation data, and their physical interrelation information in the first and second interrelation networks, it is possible to determine their logical points of alignment in the question bank space that help to clarify the degree of matching between the questions and the knowledge points.

Continuing with the above example, this step is explained in detail:

It is assumed that an intelligent question bank system is provided, which includes a plurality of target question bank entities (each entity representing a different question bank) and a plurality of knowledge point entities (each entity representing a different examination point or learning content). Corresponding embedded representation data has been generated for each of the target question bank entity and the knowledge point entity.

First embedded representation data (for question bank): for example, for a senior high school physical question bank a, a vector representing the content characteristics of the whole question bank is obtained by analyzing the stem text of all questions.

Second embedded representation data (for knowledge points): meanwhile, for a knowledge point, such as "Newton's second law", a vector representing the core concept of the knowledge point is constructed based on its definition and associated interpretation.

The question bank space logical points of the target docking node of the target question bank entity and the knowledge point entity are now determined according to the embedded representation data. This process involves the following key steps:

first, the similarity between the embedded representation of each question bank entity and the embedded representation of each knowledge point entity needs to be calculated. This can be done typically by cosine similarity, euclidean distance, etc. This similarity tells the degree of association between a question bank and a particular knowledge point.

The target docking node refers to a point where one question bank entity matches one or more knowledge point entities in the intelligent question bank system. These target docking nodes are determined by the similarity of the embedded representations calculated previously. For each question bank, the knowledge points most relevant to the question bank (i.e. the knowledge points with highest similarity) can be found, and the knowledge points become target docking nodes of the question bank.

The question bank space logical point refers to a position where the question bank and the knowledge point are butted with each other, and the position is virtual, does not refer to a physical position in the real world, but is an abstract and calculated logical position used for describing the relation between the content of the question bank and the knowledge point.

For example, if the embedded representation of the question bank a is very similar to the embedded representation of the knowledge point "newton's second law", then it can be said that in the space of the intelligent question bank system, the question bank a is very close to the logical point of the knowledge point "newton's second law". This means that the question bank a may contain many questions related to "newton's second law".

In addition, if the embedded representation of a question bank has a high similarity to the embedded representations of a plurality of knowledge points, the knowledge points are considered as target docking nodes of the question bank, and the logic positions of the question bank in the intelligent question bank system space are defined together.

By means of the method, the intelligent question bank system can understand and visualize the relation between different question banks and knowledge points, and accordingly content of the question banks is optimized, related questions are recommended to students, teaching focuses are adjusted, and the like.

Step S140, according to the question bank space logical points of each target docking node, performing a first update on the question bank space logical points of each element in the first association network and the second association network, so as to generate a first intelligent question bank frame map.

For example, a third representation element of the first association network that is combined with a second representation element of the second association network may be determined, and a fourth representation element of the second association network that is combined with the first representation element of the first association network may be determined. And then updating the first association network and the second association network until the question bank space logic point of each third expression element in the first association network and the question bank space logic point of each second expression element in the second association network are matched with the question bank space logic point of the corresponding target docking node.

For simplicity of explanation, it may be assumed that there is a specific application scenario: an intelligent question bank system on a learning platform comprises a question bank and knowledge points of mathematical subjects.

First association network (topic relation inside topic library):

assume that the mathematical problem base has the following problems:

-title a: and solving the inclined side length of the right triangle. "

-Title B: "calculate area of circle". "

-Title C: "solve equation 2x+3=11. "

These topics are presented in the form of nodes in a first association network, and links between the nodes represent logical or conceptual links between the topics. For example, topic A relates to Pythagorean theorem, topic B relates to the area formula of a circle, and topic C relates to the solution of the unitary one-time equation.

Second association network (relationship between knowledge points):

correspondingly, the knowledge point entity may be:

knowledge point 1: pythagorean theorem "

Knowledge point 2: area formula of circle "

Knowledge point 3: "unitary one-time equation solving"

In the second association network, the knowledge points are also nodes, and the connection lines between the nodes reflect the relationship between the knowledge points. For example, knowledge point 1 is related to the nature of the geometry, knowledge point 2 is also related to the measurement of the geometry, and knowledge point 3 belongs to the algebraic field.

It is first determined which topics (third presentation elements) in the first association network have a direct relationship with knowledge points (second presentation elements) in the second association network. For example, topic A is closely related to knowledge point 1 (Pythagorean theorem). At the same time, it is also determined which knowledge points (fourth representing elements) in the second association network have a direct relationship with the topics (first representing elements) in the first association network. For example, knowledge point 1 (Pythagorean theorem) is closely related to topic A.

Then, the first and second associated networks need to be updated according to these determined relationships. This means that a stronger link is established between the third representative element of topic a and the second representative element of knowledge point 1, indicating their match at the logical point in the topic library space.

The updating not only involves the matching between a single question and a knowledge point, but also involves the adjustment of each node in the whole network to ensure that the network structure can accurately reflect the actual conditions of the question bank content and knowledge system.

The updating process may require multiple iterations, each of which is adjusted according to the degree of matching of the question bank spatial logical points to the target docking node. The goal is to enable the topics in the first association network (third representation element) and the knowledge points in the second association network (fourth representation element) to be aligned as much as possible to their respective target docking nodes. This process is performed by an optimization algorithm, such as a machine learning method, which may be used to automatically adjust the connection weights between the representation elements.

After a series of updating and iteration, an optimized first intelligent question bank frame map is finally obtained. The first intelligent question bank frame map clearly shows the matching relationship among questions, knowledge points and between the questions and the knowledge points. The first intelligent question bank frame map can be used as a reference for designing courses and learning paths of students, and can also help an intelligent question bank system to more effectively conduct question recommendation, difficulty grading and learning effect evaluation.

In this way, the first intelligent question bank frame map promotes reasonable configuration of teaching resources and planning of personalized learning paths, and provides support for improving teaching and learning efficiency.

And step S150, performing second updating on the spatial logic points of the question bank of each element in the first intelligent question bank frame map according to the element interrelation information of each element in the first intelligent question bank frame map, and generating a second intelligent question bank frame map, wherein the precision of the second updating is larger than that of the first updating.

This step further refines and improves the accuracy of the first update. Re-optimizing the topic library structure using the information in the first intelligent topic library frame map may include further analyzing more subtle correlations between topics and knowledge points or introducing new data to improve accuracy of the correlation network. The resulting second version of the intelligent question bank frame map will reveal the structured relationships between the questions and knowledge points in the question bank with higher accuracy.

Illustratively, assume a first intelligent question bank framework atlas:

a question bank is provided, which contains a plurality of mathematical questions and knowledge points corresponding to the mathematical questions.

In the first intelligent question bank frame map, each question (e.g., question a, question B, question C) and knowledge point (e.g., knowledge point X, knowledge point Y, knowledge point Z) are nodes, and the links between the nodes represent the associations between them.

The first update builds the relationship between the topic and the knowledge point based on the textual matching of the topic content to the knowledge point description.

The step of performing a second update: the associated data in the first intelligent question bank frame map is analyzed, including the link strengths between the questions and the knowledge points, and their relative locations in the first intelligent question bank frame map. Next, target learning data such as student answer records, time overhead, common error types, etc. are collected. These target learning data provide a deeper hole that can help understand the complexity of the relationship between the topic and the knowledge point. Then, the logical points of the questions and knowledge points in the question bank space are recalculated using the target learning data. This may involve adjusting the link weights between the presentation elements or adding new links. For example, if topic A is actually more difficult than expected, it may need to be tied to more points of underlying knowledge, reflecting its real complexity. Next, the first intelligent question bank frame map is iteratively updated using a machine learning algorithm, each update aimed at improving the accuracy of the representation until a predetermined criterion is met. And generating a second intelligent question bank frame map according to the optimized question bank space logical points. The intelligent question bank frame map will show the relation between the questions and the knowledge points more accurately than the first intelligent question bank frame map.

Specific examples: it is assumed that in the first intelligent question bank frame map, a connection exists between the question a and the knowledge point X, but it is found that, by analyzing answer data of the student in the second update, the question a not only relates to the knowledge point X, but also needs to use the knowledge point W (precondition or more basic concept). Therefore, in the second intelligent question bank frame map, the connection between the question A and the knowledge point W is increased, and the connection strength between the question A and the knowledge point X is adjusted according to the data. In addition, it may be found that although the topic B has a strong correlation with the knowledge point Y in the first version of the map, in practice students use the knowledge point V more frequently when solving the topic. Thus, in the second version of the map, the relationship between the topic B and the knowledge point V is enhanced, and the effect of the knowledge point Y may be reconsidered.

Finally, the second intelligent question bank frame map has higher precision because of containing more data of actual use conditions, and can better guide the distribution of educational resources and the design of personalized learning paths.

Based on the steps, the butt joint nodes among the elements in the question bank are effectively determined and optimized by acquiring and processing the embedded representation data of the target question bank entity and the knowledge point entity and the associated network information. Specifically, first, the test points of the stem and knowledge point entities of the target question bank entity are utilized to construct a first association network and a second association network, and then the question bank space logic point of the target docking node is established based on embedded representation data and entity inter-association information. Thereafter, two updating operations are performed: generating a preliminary intelligent question bank frame map based on the target docking node by the first updating; and the logic points are further refined after the second updating, so that a second intelligent question bank frame map with higher precision is generated. Through the multi-stage optimization, the positioning accuracy and the recommendation quality of learning resources can be remarkably improved, so that personalized learning is promoted and the education effect is improved.

In one possible implementation, in step S110, obtaining first embedded representation data of each target question bank entity in the intelligent question bank system includes:

Step S111, obtaining first embedded representation data of each question stem in the target question bank entity, where the first embedded representation data includes: the logical branch angle of the element, the logical point information of the element in the data space, and the pointing information of the element in the data space.

For example, each step is described in terms of an intelligent question bank system on another specific learning platform:

in this example, assume that the intelligent question bank system has the following questions (target question bank entities):

the title A is "solving the diagonal side length of right triangle". "

Subject B "calculate area of circle. "

The title C "solve the equation 2x+3=11. "

For each topic, the following steps will be performed:

logical branch angle: analyzing the question type and solving method of the questions, such as that the question A belongs to the geometric question, needs to be solved by using Pythagorean theorem.

Logical point information: the location of the topic in the knowledge point space is determined, e.g. topic a corresponds to the pythagorean theorem knowledge point.

Pointing information: the knowledge domain to which the title may be directed is identified, such as title a to the geometry domain.

Through natural language processing and machine learning models, first embedded representation data may be generated for each topic, which contains the above information and can be used in subsequent matching and analysis processes.

In step S110, the obtaining the second embedded representation data of each knowledge point entity includes:

step S112, obtaining the examination point coding vector of each examination point in the knowledge point entity.

For example, next, for the knowledge point entity, the following operations will be performed:

the test point code vector: for knowledge points, such as the "Pythagorean theorem," a test point code vector may be generated, which is a mathematical expression that may represent key features of the knowledge point.

Step S113, for each examination point, invoking the docking node AI decision network to make a decision according to the examination point coding vector, and generating docking node confidence coefficient corresponding to the examination point, wherein the docking node confidence coefficient is the confidence coefficient that the examination point belongs to a docking node label.

For example, for each test point, the docking node AI decision network may be invoked using the test point code vector as input. The abutment node AI decision network may be a deep learning model whose task is to evaluate the confidence that the point belongs to a certain abutment node label, such as to determine if the "pythagorean theorem" belongs to a "geometric" label.

The docking node AI decision network outputs one or more docking node confidence levels reflecting the strength of correlation between each of the test points and the docking node labels.

Step S114, performing vectorization processing on the confidence coefficient of the docking node of the test point to generate a corresponding docking node vector, and fusing the docking node vector corresponding to each test point with the test point coding vector to generate second embedded representation data of the knowledge point entity.

And then, after vectorizing the confidence coefficient of the docking node, fusing the vectorized confidence coefficient with the test point code vector to obtain second embedded representation data of the knowledge point entity. This fused second embedded representation data contains the essential features of the test point and its strength of relationship with the different docking node labels.

Through the steps, embedded representation data of each question and knowledge point can be effectively constructed, and high-quality input data is provided for subsequent intelligent question bank frame map updating, question recommendation, difficulty grading and the like. Such embedded representations incorporate both the intrinsic properties of the topics and knowledge points themselves, and their relevance information throughout the knowledge system.

In one possible implementation, step S120 may include:

Step S121, performing frame initialization conversion on the target question bank entity and the knowledge point entity, respectively, to generate a target question bank entity frame and a knowledge point entity frame that conform to a covariate limitation, where the covariate limitation includes: after the conversion of the question bank space logical points of the elements of the knowledge point entity and the target question bank entity is executed, element vectors of the knowledge point entity and the target question bank entity are unchanged.

For example, the above examples are combined and specific scenarios are provided for explanation.

Assume that a mathematical subject library of a learning platform is provided. In this mathematical subject matter library, there are various types of mathematical problems (subject matter library entities), such as algebra, geometry, trigonometry, etc. Meanwhile, the question bank also covers related mathematical concepts and theories (knowledge point entities), such as 'Pythagorean theorem', 'area formula of circle', and the like.

In the framework initialization conversion, a standardized representation mode needs to be defined first to convert the topics and knowledge points into vector forms. For example, title a: the "right triangle-finding diagonal side" can be converted into a vector by text analysis, which contains representations of keywords and attributes associated with the topic.

During the conversion process, it must be ensured that the method used satisfies the covariate limitation. That is, the vector representations of the topics and knowledge points should not change after any transformation (e.g., rotation, translation). For example, regardless of how the question bank is rearranged, the vector of questions A and knowledge points 1: the vector of the Pythagorean theorem should remain unchanged.

Step S122, generating a first association network of the target question bank entity according to the target question bank entity frame and each of the first embedded representation data.

For example, once there is a vector representation of each topic, a network of associations between topics may be constructed from these representations. For example, if the vector of topic A is equal to topic B: the "areas of calculated circles" have a high degree of similarity, they are connected in the first associated network, since they may all involve calculation of geometry.

Step S123, obtaining a feature deviation degree between each of the test points in the knowledge point entity, if the feature deviation degree between two of the test points is smaller than a set deviation degree, determining that a connection line exists between the second representation elements corresponding to the two test points, and combining each of the second embedded representation data and the connection line between each of the second representation elements into the second association network.

For example, in a second association network of knowledge point entities, feature deviations between different points need to be evaluated. For example, knowledge point X: "Pythagorean theorem" and knowledge point Y: the area formula of a triangle may be very close in some properties (e.g., both triangle calculations) to the knowledge point Z: the degree of characteristic deviation of the "linear equation solution" is large.

If the characteristic deviation degree of the two examination points is lower than a preset threshold value, a connecting line is drawn between the two knowledge points, so that the two knowledge points are closely related in a knowledge system. For example, there will be a line between knowledge point X and knowledge point Y, as they are both related to triangles. All such combinations of links and knowledge point vectors form a second association network.

Through the above steps, the first and second association networks reflecting the complex relationship between the topics and the knowledge points can be created. These networks help reveal the inherent links between topics and knowledge points, providing a deeper hole for educators and learners.

In one possible implementation, step S121 may include:

The knowledge point entity and the target question bank entity are respectively subjected to the following steps:

step S1211, obtaining a corresponding question bank space logical point of the entity in the question bank knowledge logical space, where the source logical point of the entity is at an initial logical point of the question bank knowledge logical space.

For example, assuming a knowledge point entity "Pythagorean theorem," the source logical point of this knowledge point may be the basic concept within the geometric domain to solve the right triangle problem. Consider a target topic library entity, topic A: "calculate hypotenuse length of right triangle". Its source logical point is the type of mathematical problem it involves and the geometric knowledge required.

Step S1212, determining derived path characteristics of the derived path of the entity according to the question bank space logical points of the entity, wherein the derived path characteristics reflect the derived path information of the entity.

For example, for the knowledge point "Pythagorean theorem," the derivative path feature may include its relationship to other geometric concepts (e.g., similar triangles, trigonometric functions, etc.), as well as applications in different difficulty levels and topics. The derivative path characteristics of topic A may include the level of difficulty of the topic, its association with other mathematical concepts (e.g., algebraic expressions, linear equations, etc.), and the logical derivation steps in the solution process.

Step S1213, performing enhanced conversion processing on the entity, and optimizing the derivative path feature according to the enhanced conversion processing.

For example, the enhanced conversion process of the "Pythagorean theorem" may include optimizing its derivative path characteristics with consideration of more student answer data and historical answer success rates.

For topic A, the enhanced conversion process may involve analyzing the answering process and error patterns of more students to adjust their position in the logical space and optimize the association of topics with relevant knowledge points.

Step S1214, if the derived path information reflected by the derived path feature matches the question bank knowledge logic space, terminating the enhancement conversion process, and using the current entity frame as the initialized entity frame.

For example, if, after the enhanced transformation process, the derived path features of the "Pythagorean theorem" find a match in the topic library knowledge logic space (e.g., closely connected to a series of related geometric topics), then the further transformation process may be terminated and the current knowledge point entity frame may be determined.

Similarly, if the topic a is optimized, the position in the logical space can accurately reflect the association between the difficulty and the knowledge point, the enhanced conversion process is stopped, and the target topic library entity frame is confirmed.

Through the steps, the constructed knowledge points and the topic entity frames can accurately reflect the structures and the layers in the education field, so that a more accurate learning path is provided for learners, and a convenient teaching resource management tool is provided for teachers.

In one possible implementation, the entity-related information includes: vector cross-correlation information and logical point cross-correlation information.

For example, entity cross-correlation information contains two types of cross-correlation information: vector cross-correlation information refers to the relationship between different entity embedded representations, such as the vector of topic a and knowledge point X: the vector of the Pythagorean theorem matches or correlates to how much. The logical point cross-correlation information refers to the positional relationship between the topic and the knowledge point in the abstract knowledge space, for example, how close the topic a is to the logical point of the Pythagorean theorem in the knowledge space.

Step S130 may include: and according to the first embedded representation data and the second embedded representation data, calling a docking node matching model to execute multi-stage question bank space logic point decisions on the first representation elements and the second representation elements, and generating the question bank space logic points of the target docking node, wherein the input of the question bank space logic point decisions of each stage is the element logic point and the element vector obtained by the question bank space logic point decisions of the previous stage.

For example, the goal of this step is to find the best matching point (i.e., the target docking node) between the topic and knowledge points and determine their logical points in the topic library space. Can be decomposed into the following small steps:

a docking node matching model is invoked, which may be a complex machine learning algorithm, with the objective of finding the best docking point between the topic and the knowledge point.

The specific scene is as follows: for example, when the first embedded representation data of topic A and the second embedded representation data of knowledge point X are entered, the model evaluates which knowledge points are most relevant to topic A. This may involve calculating the similarity between vectors, taking into account the proximity of logical points, etc.

Performing multi-stage question bank space logical point decision: this process is typically iterative, with each stage optimizing decisions based on the results of the previous stage. The specific scene is as follows: in the first stage, the docking node matching model may initially determine that the association of the topic a with the pythagorean theorem is high. In the subsequent stage, the model further considers other factors, such as answer records of students, question difficulty and the like, so as to adjust the logical point relation between the questions A and the Pythagorean theorem.

Generating a question bank space logic point of the target docking node: finally, the docking node matching model outputs a logical point, which indicates that the topic A should be tightly docked with the Pythagorean theorem in the topic library space. For example, the docking node matching model may determine that the best docking point for topic A is the "hypotenuse problem" which is a more specific sub-domain under the Pythagorean theorem knowledge point. Through the matching, the intelligent question bank can provide more accurate personalized learning suggestions and improve the teaching method and learning path design of knowledge points.

Through the steps, the questions and the knowledge points can be effectively and accurately matched, and a more intelligent and highly-adaptive question bank structure is created, so that the utilization efficiency of educational resources and individuation of learning experience are improved.

In another possible implementation, which is different from the previous embodiment, the step S150 may include: and calling a docking node matching model to execute multi-stage question bank space logic point decisions on a first representing element and a second representing element which accord with a set selection rule in the first intelligent question bank frame map, and generating the question bank space logic point of the target docking node, wherein the input of the question bank space logic point decisions of each stage is an element logic point and an element vector generated by the question bank space logic point decisions of the previous stage.

The setting selection rule includes: the reference feature deviation degree between the dependent element of the first representing element and the first representing element is smaller than the set deviation degree, and the dependent element of the first representing element is a second representing element. The reference feature deviation degree between the dependent element of the second representing element and the second representing element is smaller than the set deviation degree, and the dependent element of the second representing element is the first representing element.

This process may be explained by a specific example, for example. Assume that the intelligent question bank system is attempting to update its frame map to better match the questions and knowledge points.

Assume that in a first intelligent question bank frame map, there are two elements: "hypotenuse solving" (first representing element) and "Pythagorean theorem" (second representing element). They have been initially related by some algorithm but require further validation and refinement.

Next, the docking node matching model is invoked and a multi-stage logical point decision is started. In the first stage, the docking node matching model evaluates the logical points of the "hypotenuse solving" and considers the logical points and vectors of the knowledge point "Pythagorean theorem" associated therewith.

In a subsequent stage, each decision is adjusted based on the previous result. For example, if a student is found to often refer to the "Pythagorean theorem" when solving a related topic, the association between these two elements may be enhanced.

"Right triangle hypotenuse solving" as a first representative element relies on the element being a fundamental knowledge of geometry, such as triangle properties. It may be checked whether the degree of feature deviation between these dependent elements and the first representing element is smaller than a preset threshold value. Meanwhile, the pythagorean theorem is taken as a second expression element, and a similar checking process is also adopted.

If the reference feature deviation degree between two elements is smaller than the set deviation degree and they are mutually dependent elements of each other, they accord with the selection rule, and the docking node matching model regards them as key objects in the updating process.

Through the iterative process, the logical points of the questions and the knowledge points in the first intelligent question bank frame map are finely adjusted, and finally a second intelligent question bank frame map is generated. The updated frame map reflects the relation between different questions and knowledge points more accurately, helps to improve learning efficiency, and enables education resources to be distributed more scientifically and reasonably.

In one possible implementation, the step of determining the spatial logical point of the question bank at each stage includes:

And step A110, performing excitation function processing on the element vector of the element and the element vector of each dependent element of the element to generate vector cross-correlation information of the element.

For example, assume that topic A is being processed: "calculate hypotenuse length of right triangle". First, a vector representation of this topic (element) and a vector representation of the knowledge points it depends on (e.g., pythagorean theorem) need to be obtained.

Excitation functions process these vectors, possibly involving the use of nonlinear transformations to enhance the expressive power of the features. For example, if the understanding of the Pythagorean theorem is critical to solving the topic A, this can be highlighted by the incentive function.

And step A120, performing excitation function processing on the vector cross-correlation information of the element, and fusing an excitation function processing result and element logic point deviation of the element to generate the element logic point cross-correlation information, wherein the element logic point deviation is a question bank space logic point deviation between the element and each dependent element of the element.

For example, the results of the excitation function processing may then be combined with the current logical point deviation for topic A. Logical point bias refers to the distance between the position of the topic a in the topic library space and the knowledge point on which it depends (e.g. the pythagorean theorem).

And combining the result of the excitation function and the logical point deviation, and generating the logical point cross-correlation information of the title A. This process may involve balancing the impact of different dependent elements to find the best logical point representation.

And step A130, performing feature optimization according to the element vector of the element and the vector cross-correlation information to generate an optimized element vector.

For example, the features of topic A can now be optimized based on the element vector and vector cross-correlation information. This may involve the difficulty of readjusting the topic, analyzing the topic or looking at the depth of the knowledge point, etc. The optimized element vector more accurately represents the topic A, and helps to determine the correct relationship between the element vector and other topics and knowledge points.

And step A140, performing logic point optimization according to the element logic points of the elements and the logic point interrelation information, and generating optimized element logic points.

For example, logical point optimization may be performed using the logical points of topic A and the logical point cross-correlation information. This means that the position of the topic a in the knowledge graph can be fine-tuned according to the relation between the topic and the knowledge points. The optimized logical points may bring the topic a closer to some knowledge points and farther from other less relevant knowledge points. In this way, students can be more effectively guided to related learning resources and concepts when they contact topic a.

Through the steps, the relation between the questions and the knowledge points can be dynamically adjusted, so that more customized learning experience is provided for learners, and a powerful teaching support tool is provided for teachers. Such a system can continually update and optimize itself to maintain synchronization with educational needs.

In one possible implementation, the vector cross-correlation information includes: vector optimization information in entities and vector optimization information between entities.

Step a110 may include:

And step A111, performing excitation function processing on the element vector of the element, the element vector of the first dependent element of the element and the first reference feature deviation degree between the element and the first dependent element to generate intra-entity vector information of the element, wherein the first dependent element is an element corresponding to the same entity as the element.

And step A112, fusing the intra-entity vector information corresponding to each first dependent element to generate the intra-entity vector optimization information of the element.

And step A113, performing excitation function processing on the element vector of the element, the element vector of a second dependent element of the element and a first reference characteristic deviation degree between the element and the second dependent element, and generating inter-entity vector information of the element, wherein the second dependent element is an element corresponding to a different entity from the element.

And step A114, fusing the vector information among entities corresponding to each second dependent element to generate the vector optimization information among entities of the elements.

The logical point cross-correlation information includes: logic point optimization information in the entities and logic point optimization information between the entities.

For example, for vector cross-correlation information, the description of intra-entity vector optimization information is as follows:

There is provided a number topic B (element) and its associated knowledge point Y (first dependent element), which all correspond to the same entity, i.e. "triangle area calculation".

The element vectors of the topic B and the knowledge point Y and the degree of deviation of the reference feature between them can be processed using an excitation function. For example, if the topic B is "calculate area of isosceles triangle" and the knowledge point Y is "bottom by high divided by two", then the system will analyze the vector representations of the topic text and knowledge point descriptions and generate in-vivo vector information by the excitation function taking into account their similarity or difference.

And then, the intra-entity vector information of all knowledge points related to the topic B can be fused, and more comprehensive intra-entity vector optimization information is created, so that the position and the connection of the topic B in the frame map are improved.

The description of the vector optimization information between entities is as follows:

Likewise, topic B may also depend on elements of other different entities, such as the cognitive knowledge point Z of the geometry (the second dependent element), which may be another discipline or knowledge domain.

Vector representations between the topics B and the knowledge points Z and reference feature deviations can be analyzed using the excitation function to generate inter-entity vector information.

All relevant inter-entity vector information may then be consolidated to yield inter-entity vector optimization information for topic B, which facilitates establishing a tighter connection across disciplines or across domains.

For the logical point cross-correlation information, the description of the logical point optimization information in the entity is as follows:

In the intelligent question bank, the question B has a relation with the knowledge point Y in the vector space and also has a specific position in the logic space. Therefore, the closeness of the topic B and the knowledge point Y in the logical space, i.e. the logical point association between them, needs to be considered.

By analyzing the locations of the topics and knowledge points in the logical space, logical point optimization information within the entity can be created to ensure that the topics and knowledge points are logically coherent.

The description of the logical point optimization information between entities is as follows:

For topic B and the knowledge point Z of geometry knowledge, it is also necessary to ensure that they are properly related in logical space, although they belong to different disciplines or knowledge domains.

By evaluating the interdisciplinary logical relationship between the topics and the knowledge points, the logic point optimization information between the entities can be created, so that the relevance and practicability of learning resources are improved.

In the above scenario, through careful vector and logical point optimization, the inherent relation between the topics and the knowledge points can be captured and reflected more accurately, and more effective learning paths and recommendations are provided for users.

Step a120 may include:

for each of said first dependent elements of said elements the following steps are performed:

And step A121, performing excitation function processing on the intra-entity vector information corresponding to the first dependent element to generate a first excitation eigenvalue, and performing product operation on element logic point deviation corresponding to the first dependent element and the first excitation eigenvalue to generate an intra-entity logic point eigenvalue corresponding to the first dependent element.

For example, in this example, it will be further explained how to refine the association between topics and knowledge points in the intelligent topic library system of the learning platform. The process involves performing an excitation function process on vector cross-correlation information and logical point bias of the element, and generating logical point cross-correlation information of the element by a product operation.

The specific scene is as follows: consider a specific title a: the first dependency element of "solving for the hypotenuse length of a right triangle" is the basic geometric knowledge B (e.g., the nature of the triangle).

First, excitation function processing is performed on the element B dependent interior vector information (possibly a knowledge representation of the geometrical properties) to obtain a first excitation feature value. This process may be through a neural network layer to enhance the expression of important features.

Next, the logical point bias of element B in the question bank space (i.e., its distance or correlation with the question a) is multiplied by the first excitation eigenvalue to obtain the logical point eigenvalue in the entity.

And step A122, integrating the intra-entity logic point characteristic values corresponding to the first dependent elements of the elements to generate the intra-entity logic point optimization information of the elements.

For example, for topic A, all intra-entity logical point feature values of the first dependent element (e.g., all underlying geometric knowledge points) need to be considered. The intra-entity logical point feature values are integrated to form intra-entity logical point optimization information for topic a. Integration may include methods such as weighted averaging, summary statistics, and the like.

Wherein the following steps are performed for each of the second dependent elements of the elements:

1. And performing excitation function processing on the vector information between entities corresponding to the second dependent element to generate a second excitation eigenvalue, and performing product operation on element logic point deviation corresponding to the second dependent element and the second excitation eigenvalue to generate an inter-entity logic point eigenvalue corresponding to the second dependent element.

2. And integrating the inter-entity logic point characteristic values corresponding to the second dependent elements of the elements to generate the inter-entity logic point optimization information of the elements.

For example, at the same time, topic A also has a second dependent element, such as advanced mathematical concepts C (e.g., "parsing geometry"). For these element C dependent extrinsic vector information (possibly a knowledge representation of a higher level concept) the excitation function processing is performed as well, resulting in a second excitation feature value. Then, the logical point deviation of element C (which is a logical distance from topic A) is multiplied by the second excitation eigenvalue to obtain the logical point eigenvalue between entities.

Then, the inter-entity logical point feature values of all the second dependent elements (e.g., all the high-level data concepts) can be integrated to generate inter-entity logical point optimization information for topic A.

Through the above steps, topic A considers not only the relationships of its underlying knowledge points (within entities) but also the associations with a broader mathematical domain (between entities). This detailed analysis allows the location of topic a to be more accurately located in the intelligent question bank frame map, provides students with a more accurate learning path, and allows the teacher to see how students are linked between different concepts. The optimization can promote personalized learning and finally improve the overall teaching effect.

In one possible embodiment, step a122 may include:

Step a1221, calculating an addition value of the first reference feature deviation degree corresponding to the first dependent element and 1, and calculating a first quotient between the intra-entity logical point feature value corresponding to the first dependent element and the addition value.

For example, assume that there is one mathematical topic C (element) in the intelligent topic library, such as "calculate area of equilateral triangle", which depends on several knowledge points, one of which is knowledge point P (first dependent element), i.e., "area formula of arbitrary triangle". The degree of deviation of the reference feature corresponding to the knowledge point P, which measures the degree of logical association between the topic C and the knowledge point P, can be calculated. For example, if the description of the topic C and the content of the knowledge point P are very close, the degree of deviation is low; if the phase difference is far, the degree of deviation is high. The value of the addition of the calculated degree of deviation to 1 may then be calculated, and then the quotient (first quotient value) between the logical point feature value of the knowledge point P and this added value is calculated. This operation may be to reduce the impact of those dependent elements that deviate to a large degree (weak relevance), thereby giving more attention to those knowledge points that are closely related to topic C.

Step a1222, fusing the first quotient values corresponding to the first dependent elements corresponding to the elements, and generating the logic point optimization information in the entity of the elements.

Using the topic C as an example, the topic C may depend on other knowledge points, such as the knowledge point Q (another first dependent element), e.g. "nature of an equilateral triangle", in addition to the knowledge point P.

For each dependent element of topic C (knowledge point P, Q, etc.), the system repeats the above-described step of calculating the first quotient. The system then fuses all of these quotient values in some form, possibly a weighted average or other statistical method, to generate intra-entity logical point optimization information for topic C.

By the method, the logic point optimization information in the entity of the topic C can better reflect the logic position of the knowledge point closely related to the topic C, and the topic C is ensured to be represented in the intelligent topic library more accurately and effectively.

In the whole process, the integration and fusion strategy based on the characteristic values of the element-dependent logic points is beneficial to fine adjustment of the position of the questions in the knowledge graph, so that the learning path is more personalized and targeted, and the learning effect is improved.

In one possible implementation manner, the process of integrating the inter-entity logical point feature values corresponding to the second dependent elements of the elements to generate the inter-entity logical point optimization information of the elements includes:

For each of said second dependent elements of said elements the following steps are performed:

1. And calculating an addition value of the second reference feature deviation degree corresponding to the second dependent element and 1, and calculating a second quotient between the intra-entity logic point feature value corresponding to the second dependent element and the addition value.

For example, how to perform the integration processing of the feature values of the logical points between entities on the topics (elements) in the intelligent topic library system of the learning platform will be described in detail. This process involves computing feature bias and logical point feature values for each second dependent element (knowledge points belonging to different entities) and fusing them to generate optimization information.

Illustratively, consider the title C: "solving right triangles" also relies on algebraic knowledge, such as "solving a unitary quadratic equation" (the second dependent element), in addition to geometry knowledge. We want to evaluate and optimize the association between these different entities.

For each second dependent element of topic C, such as algebraic knowledge D, a second degree of reference feature deviation between D and topic C is first determined, which measures the degree of difference between the two at logical points in the topic library space.

Next, the added value of the second reference feature deviation of D and the value 1 is calculated. Then, the added value is divided by the characteristic value of the logic point in the entity of D to obtain a second quotient. This process can be seen as adjusting the impact of the feature values of the logical points between entities by weight.

2. And fusing second quotient values corresponding to the second dependent elements corresponding to the elements to generate the logic point optimization information in the entity of the elements.

For example, a second quotient of all the second dependent elements of topic C now needs to be fused to create a comprehensive view that helps reflect the overall relationship of topic C with the cross-entity dependent knowledge points.

Fusion may involve various methods, such as weighted averaging or other statistical methods, intended to ensure that the effects of all relevant second dependent elements are properly taken into account.

Finally, the result obtained by fusion is the optimization information of the logical points in the entity of the topic C. This information can direct the intelligent question bank system to update the position of the question C in the knowledge graph, ensuring that it remains properly tied to all relevant knowledge points (whether basic knowledge or advanced knowledge).

Through the technical steps, the cross-domain knowledge points can be processed and integrated more accurately, so that a more personalized and effective learning path is provided for the user. At the same time, this optimization process also helps reveal hidden links between complex concepts, facilitating interdisciplinary learning and understanding.

Fig. 2 schematically illustrates a learning platform system 100 that may be used to implement various embodiments described in the present application.

For one embodiment, fig. 2 shows a learning platform system 100, the learning platform system 100 having a plurality of processors 102, a control module (chipset) 104 coupled to one or more of the processor(s) 102, a memory 106 coupled to the control module 104, a non-volatile memory (NVM)/storage 108 coupled to the control module 104, a plurality of input/output devices 110 coupled to the control module 104, and a network interface 112 coupled to the control module 104.

Processor 102 may include a plurality of single-core or multi-core processors, and processor 102 may include any combination of general-purpose or special-purpose processors (e.g., graphics processors, application processors, baseband processors, etc.). In some alternative embodiments, the learning platform system 100 can be used as a server device such as a gateway in the embodiments of the present application.

In some alternative embodiments, the learning platform system 100 may include a plurality of computer-readable media (e.g., memory 106 or NVM/storage 108) having instructions 114 and a plurality of processors 102 combined with the plurality of computer-readable media configured to execute the instructions 114 to implement the modules to perform the actions described in this disclosure.

For one embodiment, the control module 104 may include any suitable interface controller to provide any suitable interface to one or more of the processor(s) 102 and/or any suitable management end or component in communication with the control module 104.

The control module 104 may include a memory controller module to provide an interface to the memory 106. The memory controller modules may be hardware modules, software modules, and/or firmware modules.

Memory 106 may be used, for example, to load and store data and/or instructions 114 for learning platform system 100. For one embodiment, memory 106 may comprise any suitable volatile memory, such as, for example, a suitable DRAM. In some alternative embodiments, memory 106 may comprise a double data rate type four synchronous dynamic random access memory.

For one embodiment, the control module 104 may include a plurality of input/output controllers to provide interfaces to the NVM/storage 108 and the input/output device(s) 110.

For example, NVM/storage 108 may be used to store data and/or instructions 114. NVM/storage 108 may include any suitable non-volatile memory (e.g., flash memory) and/or may include any suitable non-volatile storage(s).

NVM/storage 108 may include storage resources that are physically part of the management side on which learning platform system 100 is installed, or which may be accessible by the device or may not be necessarily part of the device. For example, NVM/storage 108 may be accessed via input/output device(s) 110 according to a network.

The input/output device(s) 110 may provide an interface for the learning platform system 100 to communicate with any other suitable management end, and the input/output device 110 may include a communication component, a pinyin component, a sensor component, and the like. The network interface 112 may provide an interface for the learning platform system 100 to communicate in accordance with a plurality of networks, and the learning platform system 100 may communicate wirelessly with a plurality of components of a wireless network in accordance with any of a plurality of wireless network standards and/or protocols, such as accessing a wireless network in accordance with a communication standard, such as WiFi, 2G, 3G, 4G, 5G, etc., or a combination thereof.

For one embodiment, one or more of the processor(s) 102 may be packaged together with logic of a plurality of controllers (e.g., memory controller modules) of the control module 104. For one embodiment, one or more of the processor(s) 102 may be packaged together with logic of multiple controllers of the control module 104 to form a system in package. For one embodiment, one or more of the processor(s) 102 may be integrated on the same die as logic of the multiple controllers of the control module 104. For one embodiment, one or more of the processor(s) 102 may be integrated on the same die with logic of multiple controllers of the control module 104 to form a system-on-chip.

In various embodiments, the learning platform system 100 may be, but is not limited to being: a desktop computing device or a mobile computing device (e.g., a laptop computing device, a handheld computing device, a tablet, a netbook, etc.), and the like. In various embodiments, the learning platform system 100 may have more or fewer components and/or different architectures. For example, in some alternative embodiments, the learning platform system 100 includes a plurality of cameras, a keyboard, a liquid crystal display screen (including a touch screen display), a non-volatile memory port, a plurality of antennas, a graphics chip, an application specific integrated circuit, and speakers.

The foregoing has outlined rather broadly the more detailed description of the application in order that the detailed description of the principles and embodiments of the application may be implemented in conjunction with the detailed description of the application that follows, the examples being merely intended to facilitate an understanding of the method of the application and its core concepts; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present application, the present description should not be construed as limiting the present application in view of the above.

Claims

1. The title database data optimization method based on the intelligent title database system is characterized by being applied to a learning platform system, and comprises the following steps:

Acquiring first embedded representation data of a target question bank entity in an intelligent question bank system, and acquiring second embedded representation data of each knowledge point entity, wherein the first representation element in the target question bank entity is a question stem, and the second representation element in the knowledge point entity is a test point;

performing second updating on the spatial logic points of the question bank of each element in the first intelligent question bank frame map according to the element interrelation information of each element in the first intelligent question bank frame map to generate a second intelligent question bank frame map, wherein the precision of the second updating is greater than that of the first updating;

The second updating of the spatial logical points of the question bank of each element in the first intelligent question bank frame map is performed according to the element inter-related information of each element in the first intelligent question bank frame map, so as to generate a second intelligent question bank frame map, which comprises the following steps:

Invoking a docking node matching model to execute multi-stage question bank space logic point decisions on a first representing element and a second representing element which accord with a set selection rule in the first intelligent question bank frame map, and generating a question bank space logic point of the target docking node, wherein the input of the question bank space logic point decisions of each stage is an element logic point and an element vector generated by the question bank space logic point decisions of the previous stage;

the setting selection rule includes: the reference characteristic deviation degree between the dependent element of the first representing element and the first representing element is smaller than the set deviation degree, and the dependent element of the first representing element is a second representing element; the reference feature deviation degree between the dependent element of the second representing element and the second representing element is smaller than the set deviation degree, and the dependent element of the second representing element is the first representing element;

The step of the spatial logical point decision of the question bank in each stage comprises the following steps:

Performing excitation function processing on the element vector of the element and the element vector of each dependent element of the element to generate vector cross-correlation information of the element;

Performing excitation function processing on the vector cross-correlation information of the element, and fusing an excitation function processing result and element logic point deviation of the element to generate the element logic point cross-correlation information, wherein the element logic point deviation is a question bank space logic point deviation between the element and each dependent element of the element;

performing feature optimization according to the element vector of the element and the vector cross-correlation information to generate an optimized element vector;

performing logical point optimization according to the element logical points of the elements and the logical point inter-related information to generate optimized element logical points;

The vector cross-correlation information includes: vector optimization information in entities and vector optimization information between entities;

The excitation function processing is performed on the element vector of the element and the element vector of each dependent element of the element, and vector cross-correlation information of the element is generated, and the method comprises the following steps:

Performing excitation function processing on the element vector of the element, the element vector of a first dependent element of the element, and a first reference feature deviation degree between the element and the first dependent element, so as to generate intra-entity vector information of the element, wherein the first dependent element is an element corresponding to the same entity as the element;

fusing the intra-entity vector information corresponding to each first dependent element to generate intra-entity vector optimization information of the element;

Performing excitation function processing on the element vector of the element, the element vector of a second dependent element of the element, and a first reference feature deviation degree between the element and the second dependent element, so as to generate inter-entity vector information of the element, wherein the second dependent element is an element corresponding to a different entity from the element;

Fusing the vector information among entities corresponding to each second dependent element to generate vector optimization information among entities of the elements;

The logical point cross-correlation information includes: logic point optimization information in the entities and logic point optimization information between the entities;

The vector cross-correlation information of the element is subjected to excitation function processing, and an excitation function processing result and element logic point deviation of the element are fused to generate the element logic point cross-correlation information, which comprises the following steps:

performing excitation function processing on the intra-entity vector information corresponding to the first dependent element to generate a first excitation characteristic value, and performing product operation on element logic point deviation corresponding to the first dependent element and the first excitation characteristic value to generate an intra-entity logic point characteristic value corresponding to the first dependent element;

Integrating the intra-entity logic point characteristic values corresponding to the first dependent elements of the elements to generate intra-entity logic point optimization information of the elements;

Performing excitation function processing on the vector information between entities corresponding to the second dependent element to generate a second excitation characteristic value, and performing product operation on element logic point deviation corresponding to the second dependent element and the second excitation characteristic value to generate an inter-entity logic point characteristic value corresponding to the second dependent element;

Integrating the inter-entity logic point characteristic values corresponding to the second dependent elements of the elements to generate the inter-entity logic point optimization information of the elements;

Integrating the intra-entity logic point characteristic values corresponding to the first dependent elements of the elements to generate the intra-entity logic point optimization information of the elements, wherein the method comprises the following steps:

calculating an addition value of a first reference feature deviation degree corresponding to the first dependent element and 1, and calculating a first quotient between the intra-entity logic point feature value corresponding to the first dependent element and the addition value;

fusing first quotient values corresponding to the first dependent elements corresponding to the elements to generate logic point optimization information in the entity of the elements;

The integrating the inter-entity logical point feature values corresponding to the second dependent elements of the elements to generate the inter-entity logical point optimization information of the elements includes:

Calculating an addition value of a second reference feature deviation degree corresponding to the second dependent element and 1, and calculating a second quotient between the intra-entity logic point feature value corresponding to the second dependent element and the addition value;

Fusing second quotient values corresponding to the second dependent elements corresponding to the elements to generate logic point optimization information in the entity of the elements;

The first updating of the spatial logical points of the question bank of each element in the first association network and the second association network is performed according to the spatial logical points of the question bank of each target docking node, and a first intelligent question bank frame map is generated, which comprises:

Determining a third representation element in the first association network that is combined with a second representation element in the second association network, and determining a fourth representation element in the second association network that is combined with the first representation element in the first association network;

Updating the first association network and the second association network until the question bank space logic point of each third expression element in the first association network and the question bank space logic point of each second expression element in the second association network are matched with the question bank space logic point of the corresponding target docking node.

2. The method for optimizing question bank data based on the intelligent question bank system according to claim 1, wherein the obtaining the first embedded representation data of each target question bank entity in the intelligent question bank system comprises:

Acquiring first embedded representation data of each question stem in the target question bank entity, wherein the first embedded representation data comprises: the logical branch angle of the element, the logical point information of the element in the data space and the pointing information of the element in the data space;

The obtaining the second embedded representation data of each knowledge point entity comprises:

Acquiring a test point code vector of each test point in the knowledge point entity;

For each examination point, invoking the docking node AI decision network to make a decision according to the examination point coding vector, and generating docking node confidence coefficient corresponding to the examination point, wherein the docking node confidence coefficient is the confidence coefficient that the examination point belongs to a docking node label;

And carrying out vectorization processing on the confidence coefficient of the docking node of the test point to generate a corresponding docking node vector, and fusing the docking node vector corresponding to each test point with the test point coding vector to generate second embedded representation data of the knowledge point entity.

3. The method of claim 1, wherein the extracting the first association network of the target question bank entity and the second association network of the knowledge point entity comprises:

Respectively carrying out frame initialization conversion on the target question bank entity and the knowledge point entity to generate a target question bank entity frame and a knowledge point entity frame which accord with covariate limitation, wherein the covariate limitation comprises: after the conversion of the question bank space logical points of the elements of the knowledge point entity and the target question bank entity is executed, the element vectors of the knowledge point entity and the target question bank entity are unchanged;

Generating a first association network of the target question bank entity according to the target question bank entity framework and each first embedded representation data;

And acquiring the characteristic deviation degree between the examination points in the knowledge point entity, if the characteristic deviation degree between the two examination points is smaller than the set deviation degree, determining that a connecting line exists between the second representation elements corresponding to the two examination points, and combining the second embedded representation data and the connecting line between the second representation elements into the second association network.

4. The method for optimizing question bank data based on intelligent question bank system according to claim 3, wherein the step of performing frame initialization conversion on the target question bank entity and the knowledge point entity respectively to generate a target question bank entity frame and a knowledge point entity frame which meet covariate limitation comprises the steps of:

acquiring a corresponding question bank space logic point of the entity in a question bank knowledge logic space, wherein a source logic point of the entity is positioned at an initial logic point of the question bank knowledge logic space;

Determining derivative path characteristics of the derivative path of the entity according to the question bank space logical points of the entity, wherein the derivative path characteristics reflect the derivative path information of the entity;

Performing enhancement conversion processing on the entity, and optimizing the derivative path characteristics according to the enhancement conversion processing;

and if the derived path information reflected by the derived path characteristics is matched with the question bank knowledge logic space, terminating the enhancement conversion processing, and taking the current entity frame as an initialized entity frame.

5. The method for optimizing question bank data based on an intelligent question bank system according to claim 1, wherein the entity-related information comprises: vector cross-correlation information and logical point cross-correlation information; the determining the question bank space logical point of the target butt joint node of the target question bank entity and the knowledge point entity according to the first embedded representation data, the second embedded representation data and the entity mutual association information comprises the following steps:

And according to the first embedded representation data and the second embedded representation data, calling a docking node matching model to execute multi-stage question bank space logic point decisions on the first representation elements and the second representation elements, and generating the question bank space logic points of the target docking node, wherein the input of the question bank space logic point decisions of each stage is the element logic point and the element vector obtained by the question bank space logic point decisions of the previous stage.

6. A learning platform system comprising a processor and a machine-readable storage medium having stored therein machine-executable instructions loaded and executed by the processor to implement the intelligent question bank system-based question bank data optimization method of any of claims 1-5.