An intelligent question generation method for knowledge point analysis

Abstract

The present invention relates to the technical field of question generation, and discloses an intelligent question generation method for knowledge point analysis. The method includes: performing knowledge point annotation and entity annotation on math questions, and clustering math questions that examine the same knowledge based on the entity annotation results; extracting templates from the math question clusters by using the association rule template mining method; using a variable entity content generation model to generate the variable entity content in the question template and the assignment results of the immutable entities, and filling them into the question template to form questions. The present invention uses the entity annotation results to merge math questions with consistent knowledge examined in the questions, similar involved theorem principles, and similar question structures into the same math question cluster, realizes entity-based math question clustering, uses the variable entity content generation model to assign values to the immutable entities, semantically iterate the variable entity content of the variable entities, obtains variable entity content with consistent context semantics, and generates questions.

Description

Technical Field

[0001] The present invention relates to the technical field of question generation, and particularly to an intelligent question generation method for knowledge point analysis. Background Art

[0002] With the rapid development of information technology, the education field is also undergoing profound changes. Especially in mathematics education, traditional teaching methods are gradually unable to meet the needs of students' personalized learning. In traditional mathematics education, teachers usually rely on textbooks and existing question banks to design questions. This method has many limitations. Firstly, the singleness and fixity of questions make it difficult to adapt to the learning progress and abilities of different students. Secondly, teachers often lack in-depth analysis of knowledge points when designing questions, resulting in unsatisfactory learning effects. Therefore, there is an urgent need for an intelligent way to generate mathematics questions, so as to make education more personalized and efficient. Summary of the Invention

[0003] In view of this, the present invention proposes an intelligent question generation method for knowledge point analysis. By analyzing the occurrence frequencies of principles and theorems in different knowledge points, extracting question templates for different knowledge points, using natural language processing methods to generate question descriptions in the question templates, and generating mathematics questions that meet the needs of knowledge education through intelligent means to improve learning effects.

[0004] To achieve the above object, an intelligent question generation method for knowledge point analysis provided by the present invention includes the following steps:

[0005] S1: Obtain mathematics questions, perform knowledge point annotation and entity annotation on the mathematics questions, and perform clustering processing on the mathematics questions examining the same knowledge based on the entity annotation results to obtain multiple mathematics question clusters examining the same knowledge, where the clustering processing uses an entity-based question structure clustering method.

[0006] S2: Use the association rule template mining method to extract templates from the mathematics question clusters to obtain the question templates of the mathematics question clusters, and annotate the entity variables in the question templates to obtain the variable entities and immutable entities in the question templates.

[0007] S3: Construct a variable entity content generation model, and use the variable entity content generation model to generate the variable entity content in the question template and the assignment results of the immutable entities. The variable entity content generation model takes the context immutable entities of the variable entities in the question template as input and outputs the variable entity content and the assignment results of the immutable entities, where the variable entity content generation model is implemented using a context-based sequence-to-sequence method.

[0008] S4: Fill the variable entity content and the assignment results of the immutable entities into the question template to form questions, and obtain the question generation results for different knowledge points.

[0009] As a further improvement method of the present invention:

[0010] Optionally, in the S1 step, obtain math questions and perform knowledge point annotation and entity annotation on the math questions, including:

[0011] Obtain N groups of math questions, and perform knowledge point annotation and entity annotation on each group of math questions. The knowledge point annotation and entity annotation process for the nth group of math questions is as follows:

[0012] Annotate the knowledge examined by the nth group of math questions. The knowledge that math questions can examine includes geometry, algebra, number theory, mathematical analysis, and probability and statistics. .

[0013] Annotate the theorems and principles involved in the nth group of math questions. Combine the knowledge annotation results, theorem, and principle annotation results to form the knowledge point annotation results of the nth group of math questions. :

[0014] .

[0015] Where:

[0016] represents the knowledge point annotation results of the nth group of math questions. represents the knowledge examined by the nth group of math questions. represents the set of theorems and principles involved in the nth group of math questions.

[0017] Identify the key elements in the nth group of math questions. The key element types of the key elements include numbers, variables, geometric shapes, operators, and units. Annotate the identified key elements and the types corresponding to the key elements to obtain the entity annotation results of the nth group of math questions. :

[0018] .

[0019] Where:

[0020] represents the cth key element in the nth group of math questions. represents the key element type. represents the total number of key elements in the nth group of math questions. , the key element types 1-5 are number, variable, geometric shape, operator, and unit in sequence; in the embodiments of the present invention, if , then the key element appears at a position earlier than the key element in the nth group of math problems.

[0021] All math problems with the same knowledge to be examined form the math problem set of this knowledge, where the math problem set of the ith knowledge is :

[0022] ;

[0023] .

[0024] Wherein:

[0025] represents the sth group of math problems examining the ith knowledge, represents the total number of math problems examining the ith knowledge;

[0026] represents the set of theorems and principles involved;

[0027] represents the entity annotation result of;

[0028] represents the cth key element of, represents the key element the key element type of, represents the number of key elements of.

[0029] Optionally, clustering the math problems examining the same knowledge based on the entity annotation result to obtain multiple math problem clusters examining the same knowledge, including:

[0030] The clustering process of the math problems examining the ith knowledge is as follows:

[0031] S11: Calculate the distance between any two groups of math problems in the math problem set , where the distance between the math problem and is:

[0032] .

[0033] Wherein:

[0034] Represents a mathematical problem and the distance between ;

[0035] Represents a mathematical problem and the distance between the key elements in

[0036] Represents the set of theorems and principles and the intersection between Represents the set of theorems and principles and the union between Represents the set of theorems and principles involved;

[0037] Represents the number of theorems and principles in the set;

[0038] Represents the entity annotation result of

[0039] Represents the exponential function with the natural constant as the base.

[0040] S12: Based on the distance between mathematical problems, calculate the clustering center information weight of each group of mathematical problems, where the clustering center information weight of the mathematical problem is :

[0041] ;

[0042] ;

[0043] .

[0044] Among them:

[0045] is the density of the mathematical problem ;

[0046] Represents the set of mathematical problems in which the number of mathematical problems whose distance from the mathematical problem is less than the preset distance threshold;

[0047] Represents the set of mathematical problems in which the number of mathematical problems whose distance from the mathematical problem A set of math problems with a distance between them less than a preset distance threshold, , denotes the set of math problems in which any math problem, denotes the math problem distance between;

[0048] denotes the math problem cluster center coefficient;

[0049] denotes the set of math problems with a density greater than in the math problems;

[0050] denotes the preset maximum cluster center coefficient.

[0051] S13: Select the math problems with the cluster center information weight higher than the preset threshold as the math problem cluster centers, and merge and cluster the math problems that examine the non-math problem cluster centers of the i-th knowledge with the nearest math problem cluster centers to form math problem clusters, where the set of math problem clusters that examine the i-th knowledge is :

[0052] .

[0053] Where:

[0054] denotes the k-th math problem cluster that examines the i-th knowledge, denotes the total number of math problem clusters that examine the i-th knowledge.

[0055] Optionally, in the S2 step, the association rule template mining method is used to extract templates from the math problem clusters, including:

[0056] Using the association rule template mining method to extract templates from the math problem clusters to obtain multiple problem templates of the math problem clusters, where the template extraction process of the math problem cluster is as follows:

[0057] S21: Perform word segmentation and stop word removal on the math problems in the math problem cluster to obtain the phrase sequence of the math problems, where the phrase sequence contains phrases with entities marked as key elements and phrases that are not key elements. The phrase sequence of the m-th group of math problems in the math problem cluster is :

[0058] .

[0059] Wherein:

[0060] represents the h-th phrase in the phrase sequence ; , represents the total number of phrases in the phrase sequence ;

[0061] , represents the number of math problems in the math problem cluster ;

[0062] If , then the phrase is a key element; if , then the phrase is not a key element

[0063] S22: Replace the phrases that are key elements in the phrase sequence with the key element type to obtain the phrase sequence after key element replacement

[0064] S23: Form a phrase set from multiple phrases in the phrase sequence, and calculate the support degree of each phrase set, where the minimum number of key element types in each phrase set is , the upper limit of the number of phrases in the phrase set is , the lower limit is , is the preset phrase set threshold, and the support degree of the phrase set Q is :

[0065] ;

[0066] .

[0067] Wherein:

[0068] represents U phrases in the phrase set Q , and U represents the total number of phrases in the phrase set Q

[0069] represents the number of math problems in the math problem cluster that contain the phrase , and the order of phrase appearance is the same as that of the phrase set Q

[0070] S24: Use the phrase set with a support degree higher than the preset support degree threshold as the problem template of the math problem cluster , and use the phrases in the problem template as entity variables

[0071] Optionally, the annotation of the entity variables in the question template to obtain the variable entities and immutable entities in the question template includes:

[0072] Annotate the entity variables in the question template that are of the key element type as immutable entities;

[0073] Annotate the entity variables in the question template that are not of the key element type as variable entities.

[0074] Optionally, the construction of the variable entity content generation model in step S3 includes:

[0075] Construct a variable entity content generation model, which takes the question template as input and outputs the variable entity content and the assignment results of the immutable entities. The variable entity content generation model is implemented using a sequence-to-sequence method based on context content. The variable entity content generation model includes an input layer, a context extraction layer, an iterative generation layer, and an output layer. The input layer is used to receive the question template and the entity variable annotation results. The context extraction layer is used to extract the context immutable entities of the variable entities in the question template and assign values to the context immutable entities. The iterative generation layer iteratively generates the variable entity content based on the context immutable entities to obtain the variable entity content of the variable entities in the question template. The output layer outputs the assignment results of the immutable entities and the variable entity content.

[0076] Optionally, the use of the variable entity content generation model to generate the variable entity content and the assignment results of the immutable entities in the question template includes:

[0077] S31: The input layer receives the question template B and the entity variable annotation results :

[0078] ;

[0079] .

[0080] Where:

[0081] represents the v-th entity variable in the question template B, , represents the total number of entity variables in the question template B;

[0082] is the annotation result of the entity variable , represents that the entity variable is a variable entity, represents that the entity variable is an immutable entity.

[0083] S32: The context extraction layer extracts the context immutable entities of the variable entities in the question template B. The context immutable entities of the variable entities include the upstream immutable entities and the downstream immutable entities. The extraction methods for the upstream immutable entities and the downstream immutable entities are as follows: obtain the upstream nearest immutable entity in the question template that is located before the variable entity and is the closest to the variable entity, and the downstream nearest immutable entity in the question template that is located after the variable entity and is the closest to the variable entity. Take the entity sequence between the variable entity and the upstream nearest immutable entity as the upstream immutable entity of the variable entity, and take the entity sequence between the variable entity and the downstream nearest immutable entity as the downstream immutable entity of the variable entity.

[0084] For the variable entity , then the upstream immutable entity is , and the downstream immutable entity is :

[0085] ;

[0086] .

[0087] Wherein:

[0088] represents the upstream nearest immutable entity of represents the downstream nearest immutable entity of

[0089] S33: Assign values to the context immutable entities. The assignment methods for the upstream immutable entity and the downstream immutable entity are as follows:

[0090] Obtain the key element type of , generate a phrase that conforms to this key element type, and assign values to to obtain the assigned upstream immutable entity

[0091] Obtain the key element type of , generate a phrase that conforms to this key element type, and assign values to to obtain the assigned downstream immutable entity

[0092] In the embodiments of the present invention, any upstream immutable entity or downstream immutable entity will only be assigned values once, and will not be reassigned until a set of questions is generated.

[0093] S34: The iterative generation layer iteratively generates the variable entity content based on the context-immutable entity, and obtains the variable entity content of the variable entity in the question template, where the variable entity The iterative generation process of the variable entity content is as follows:

[0094] S341: For and perform word vector representation to obtain a word vector sequence ; In the embodiment of the present invention, the way of word vector representation is the word2vec model.

[0095] S342: Set the current iteration number of the word vector sequence corresponding to the variable entity content to t, the initial value of t is 0, and the maximum value is Max. Then the t-th iteration result of the word vector sequence corresponding to the variable entity content of the variable entity is , , , represent the upper context word vector sequence and the lower context word vector sequence obtained in the t-th iteration in sequence.

[0096] S343: Perform semantic iterative processing with random perturbations added to the word vector sequence corresponding to the variable entity content. The semantic iterative processing formula is:

[0097] .

[0098] Where:

[0099] T represents transpose;

[0100] represents the semantic convolution matrix;

[0101] represents the convolution operator;

[0102] represents the ReLU activation function;

[0103] represents the random noise sequence;

[0104] represents the identity matrix.

[0105] Semantically iteratively process the word vector sequence corresponding to the variable entity content in sequence according to the order of the variable entity content in the question template, and the word vectors that have completed semantic iterative processing overwrite the word vectors that have not completed semantic iterative processing.

[0106] S344: Let t = t + 1, return to step S343 until the maximum number of iterations is reached, and use the word vector sequence obtained by semantic iterative processing at this time Convert it into a sequence of phrases, and use the last phrase in the sequence of phrases corresponding to the above word vector sequence as the variable entity content of the variable entity

[0107] S35: Extract the assignment results of all immutable entities in step S33 and the variable entity content of all variable entities in step S34. The output layer outputs the assignment results of the immutable entities and the variable entity content as a set of question template contents to be filled, and returns to step S33 to re-assign values to the immutable entities to generate the question template contents to be filled

[0108] Optionally, filling the variable entity content and the assignment results of the immutable entities into the question template to form questions includes:

[0109] Fill the variable entity content generated by the variable entity content generation model and the assignment results of the immutable entities into the corresponding question template to form a set of questions, and repeat this operation to generate question generation results under different knowledge and different question templates

[0110] Optionally The calculation process is:

[0111] Initialize , where , , represents the number of key elements of the math question ;

[0112] For perform recursion, where represents the distance between the first key elements of the math question and the first key elements of the math question . The recursion formula is:

[0113] .

[0114] Among them:

[0115] represents the key element type of the th key element of the math question ;

[0116] represents the key element type of the th key element of the math question ;

[0117] Indicates selection of the maximum value in

[0118] Based on the distance obtained by recursion, calculate the mathematical problem and the distances between the key elements in

[0119] .

[0120] Wherein:

[0121] represents the distance between the key elements in the mathematical problem and in

[0122] To solve the above problems, the present invention provides an electronic device, which includes:

[0123] A memory for storing at least one instruction;

[0124] A communication interface for enabling communication of the electronic device; and

[0125] A processor for executing the instructions stored in the memory to implement the intelligent problem generation method for knowledge point analysis described above.

[0126] To solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored, and the at least one instruction is executed by a processor in an electronic device to implement the intelligent problem generation method for knowledge point analysis described above.

[0127] Compared with the prior art, the present invention proposes an intelligent problem generation method for knowledge point analysis, and this technology has the following advantages:

[0128] First of all, this solution proposes a method for extracting problem templates, obtains a large number of mathematical problems, and performs knowledge point annotation and entity annotation respectively. Mark the phrases involving numbers, variables, geometric shapes, operators, and units in the mathematical problems as key elements, as the entity annotation results. Use the entity annotation results to perform entity-based problem structure clustering on the mathematical problems, and merge the mathematical problems with consistent problem knowledge, similar involved theorem principles, and similar problem structures into the same mathematical problem cluster to achieve entity-based mathematical problem clustering. Use the association rule template mining method to extract templates from the mathematical problem cluster, and extract a set of mathematical problem phrases with a support degree higher than the threshold as the problem template to achieve the extraction of problem templates for different knowledge.

[0129] Meanwhile, this solution eliminates a question generation method. It takes the question template phrases that are key elements as immutable entities, and other phrases as mutable entities. It uses the mutable entity content generation model to assign values to the immutable entities in combination with the key element types of the immutable entities, and performs semantic iteration on the mutable entity content of the mutable entities to obtain mutable entity content with consistent context semantics. Then, it fills the mutable entity content and the assignment results of the immutable entities into the question template to form questions, realizing the automatic generation of questions for different knowledge points. BRIEF DESCRIPTION OF THE DRAWINGS

[0130] Figure 1 It is a schematic flowchart of an intelligent question generation method for knowledge point analysis provided by an embodiment of the present invention.

[0131] The realization, functional features and advantages of the purpose of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0132] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0133] An embodiment of the present application provides an intelligent question generation method for knowledge point analysis. The execution subject of the intelligent question generation method for knowledge point analysis includes, but is not limited to, at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the intelligent question generation method for knowledge point analysis can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0134] Embodiment 1:

[0135] An intelligent question generation method for knowledge point analysis includes the following steps:

[0136] S1: Obtain math questions, perform knowledge point annotation and entity annotation on the math questions, and perform clustering processing on the math questions that examine the same knowledge based on the entity annotation results to obtain multiple math question clusters that examine the same knowledge.

[0137] In the S1 step, obtaining math questions and performing knowledge point annotation and entity annotation on the math questions includes:

[0138] Obtain N groups of math questions, and perform knowledge point annotation and entity annotation on each group of math questions. The knowledge point annotation and entity annotation process for the nth group of math questions is as follows:

[0139] Annotate the knowledge examined by the nth group of math problems, where the knowledge that math problems can examine includes geometry, algebra, number theory, mathematical analysis, and probability and statistics. 。

[0140] Annotate the theorems and principles involved in the nth group of math problems, and form the knowledge point annotation result of the nth group of math problems with the knowledge annotation result, theorem, and principle annotation result. :

[0141] 。

[0142] Among them:

[0143] represents the knowledge point annotation result of the nth group of math problems. represents the knowledge examined by the nth group of math problems. represents the set of theorems and principles involved in the nth group of math problems.

[0144] Identify the key elements in the nth group of math problems. The key element types of key elements include numbers, variables, geometric shapes, operators, and units. Annotate the identified key elements and the types corresponding to the key elements to obtain the entity annotation result of the nth group of math problems. :

[0145] 。

[0146] Among them:

[0147] represents the cth key element in the nth group of math problems. represents the key element. type. represents the total number of key elements in the nth group of math problems. , and the key element types 1-5 are numbers, variables, geometric shapes, operators, and units in sequence; in the embodiments of the present invention, if , then the key element appears at a position earlier than the key element in the nth group of math problems.

[0148] Form the math problem set of this knowledge by combining all math problems with the same examined knowledge. The math problem set of the ith knowledge is :

[0149] ;

[0150] 。

[0151] Among them:

[0152] represents the s-th group of math problems for examining the i-th type of knowledge, represents the total number of math problems for examining the i-th type of knowledge;

[0153] represents the set of theorems and principles involved;

[0154] represents the entity annotation result of;

[0155] represents the c-th key element of, represents the key element the key element type of, represents the number of key elements of.

[0156] Performing clustering processing on the math problems for examining the same knowledge based on the entity annotation result to obtain multiple math problem clusters for examining the same knowledge, including:

[0157] The clustering processing flow of the math problems for examining the i-th type of knowledge is as follows:

[0158] S11: Calculate the distance between any two groups of math problems in the math problem set where the distance between the math problem and is:

[0159] .

[0160] Among them:

[0161] represents the distance between the math problem and , ;

[0162] represents the distance between the key elements in the math problem and ;

[0163] represents the intersection between the set of theorems and principles and , represents the union between the set of theorems and principles and , represents Set of theorems and principles involved;

[0164] Indicates the number of theorems and principles in the set;

[0165] Indicates Entity annotation result;

[0166] Indicates the exponential function with the natural constant as the base.

[0167] S12: Based on the distance between math problems, calculate the clustering center information weight for each group of math problems, where the math problem The clustering center information weight is :

[0168] ;

[0169] ;

[0170] .

[0171] Among them:

[0172] Is the density of the math problem ;

[0173] Indicates the set of math problems In which the number of math problems with a distance less than the preset distance threshold from the math problem ;

[0174] Indicates the set of math problems In which the number of math problems with a distance less than the preset distance threshold from the math problem ; , Indicates any math problem in the set of math problems ; Indicates the math problem Distance between;

[0175] Indicates the math problem Clustering center coefficient;

[0176] Indicates the set of math problems In which the density is greater than Math problems;

[0177] Indicates the preset maximum clustering center coefficient.

[0178] S13: Select the math problems with the weight of clustering center information higher than the preset threshold as the math problem cluster centers, and merge and cluster the math problems that examine the i-th kind of knowledge and are non-math problem cluster centers with the nearest math problem cluster centers to form math problem clusters, where the set of math problem clusters that examine the i-th kind of knowledge is :

[0179] 。

[0180] Among them:

[0181] represents the k-th math problem cluster that examines the i-th kind of knowledge, represents the total number of math problem clusters that examine the i-th kind of knowledge.

[0182] S2: Use the association rule template mining method to extract templates from the math problem clusters, obtain the problem templates of the math problem clusters, and label the entity variables in the problem templates to obtain the variable entities and immutable entities in the problem templates.

[0183] In the step S2, using the association rule template mining method to extract templates from the math problem clusters includes:

[0184] Using the association rule template mining method to extract templates from the math problem clusters, obtaining multiple problem templates of the math problem clusters, where the math problem clusters The template extraction process is:

[0185] S21: Perform word segmentation and stop word removal processing on the math problems in the math problem clusters to obtain the phrase sequence of the math problems, where the phrase sequence includes phrases with entity annotations as key elements and phrases that are not key elements. The phrase sequence of the m-th group of math problems in the math problem clusters is :

[0186] 。

[0187] Among them:

[0188] represents the h-th phrase in the phrase sequence , , represents the phrase sequence The total number of phrases in;

[0189] , represents the math problem clusters The number of math problems in;

[0190] If , then the phrase is a key element. If , then the phrase is not a key element.

[0191] S22: Replace the phrases that are key elements in the phrase sequence with the key element types to obtain the phrase sequence after key element replacement.

[0192] S23: Form phrase sets from multiple phrases in the phrase sequence, and calculate the support degree of each phrase set. The minimum number of key element types in each phrase set is , the upper limit of the number of phrases in the phrase set is , the lower limit is , is the preset phrase set threshold. The support degree of the phrase set Q is :

[0193] ;

[0194] .

[0195] Among them:

[0196] represents U phrases in the phrase set Q, , and U represents the total number of phrases in the phrase set Q;

[0197] represents the mathematics problem cluster in which the mathematics problems containing the phrase , and the order of the phrase appearance is the same as that of the phrase set Q;

[0198] Mark the entity variables that are of the key element type in the problem template as immutable entities;

[0199] Mark the entity variables that are not of the key element type in the problem template as mutable entities.

[0200] S3: Build a mutable entity content generation model, and use the mutable entity content generation model to generate the mutable entity content in the problem template and the assignment results of the immutable entities.

[0201] In the step S3 of building the mutable entity content generation model, it includes:

[0202] Construct a variable entity content generation model, which takes a question template as input and outputs variable entity content and the assignment results of immutable entities. The variable entity content generation model is implemented using a sequence-to-sequence method based on context content. The variable entity content generation model includes an input layer, a context extraction layer, an iterative generation layer, and an output layer. The input layer is used to receive the question template and the entity variable annotation results. The context extraction layer is used to extract the context immutable entities of the variable entities in the question template and assign values to the context immutable entities. The iterative generation layer iteratively generates the variable entity content of the variable entities based on the context immutable entities to obtain the variable entity content of the variable entities in the question template. The output layer outputs the assignment results of the immutable entities and the variable entity content.

[0203] Generating the variable entity content and the assignment results of the immutable entities in the question template by using the variable entity content generation model includes:

[0204] S31: The input layer receives the question template B and the entity variable annotation results :

[0205] ;

[0206] .

[0207] Where:

[0208] represents the v-th entity variable in the question template B, , represents the total number of entity variables in the question template B;

[0209] is the annotation result of the entity variable , represents that the entity variable is a variable entity, represents that the entity variable is an immutable entity.

[0210] S32: The context extraction layer extracts the context immutable entities of the variable entities in the question template B. The context immutable entities of the variable entities include the above immutable entities and the below immutable entities. The extraction methods for the above immutable entities and the below immutable entities are as follows: obtain the nearest above immutable entity in the question template that is before the variable entity and closest to the variable entity, and the nearest below immutable entity in the question template that is after the variable entity and closest to the variable entity. The entity sequence between the variable entity and the nearest above immutable entity is used as the above immutable entity of the variable entity, and the entity sequence between the variable entity and the nearest below immutable entity is used as the below immutable entity of the variable entity.

[0211] For the variable entity , then the above immutable entity is , and the below immutable entity is :

[0212] ;

[0213] .

[0214] Wherein:

[0215] represents the nearest above immutable entity of represents the nearest below immutable entity of

[0216] S33: Assign values to the context immutable entities. The assignment methods for the above immutable entity and the below immutable entity are as follows:

[0217] Obtain the key element type of , generate a phrase that conforms to this key element type, and assign a value to to obtain the assigned above immutable entity

[0218] Obtain the key element type of , generate a phrase that conforms to this key element type, and assign a value to to obtain the assigned below immutable entity

[0219] In the embodiments of the present invention, any above immutable entity or below immutable entity will only be assigned a value once, and will not be reassigned until a set of questions is generated.

[0220] S34: The iterative generation layer iteratively generates the variable entity content based on the context-immutable entities to obtain the variable entity content of the variable entities in the question template, where the variable entities The iterative generation process of the variable entity content is as follows:

[0221] S341: For and perform word vector representation to obtain a word vector sequence ; In the embodiments of the present invention, the way of word vector representation is the word2vec model.

[0222] S342: Set the current iteration number of the word vector sequence corresponding to the variable entity content to t, the initial value of t is 0, and the maximum value is Max. Then the t-th iteration result of the word vector sequence corresponding to the variable entity content of the variable entity is , , , represent the upper context word vector sequence and the lower context word vector sequence obtained in the t-th iteration, respectively.

[0223] S343: Perform semantic iterative processing with random perturbations added to the word vector sequence corresponding to the variable entity content. The semantic iterative processing formula is:

[0224] .

[0225] Where:

[0226] T represents transpose;

[0227] represents the semantic convolution matrix;

[0228] represents the convolution operator;

[0229] represents the ReLU activation function;

[0230] represents the random noise sequence;

[0231] represents the identity matrix.

[0232] Semantically iteratively process the word vector sequence corresponding to the variable entity content in sequence according to the order of the variable entity content in the question template, and the word vectors that have completed semantic iterative processing overwrite the word vectors that have not completed semantic iterative processing currently.

[0233] S344: Let t = t + 1, return to step S343 until the maximum iteration number is reached, and use the word vector sequence obtained by semantic iterative processing at this time Convert it into a sequence of phrases, and use the last phrase in the sequence of phrases corresponding to the above-mentioned word vector sequence as the variable entity of the variable entity content.

[0234] S35: Extract the assignment results of all immutable entities in step S33 and the variable entity content of all variable entities in step S34. The output layer outputs the assignment results of the immutable entities and the variable entity content as a set of question template contents to be filled, and returns to step S33 to re-assign values to the immutable entities to generate the question template contents to be filled.

[0235] S4: Fill the variable entity content and the assignment results of the immutable entities into the question template to form questions, and obtain the question generation results for different knowledge points.

[0236] In step S4, filling the variable entity content and the assignment results of the immutable entities into the question template to form questions includes:

[0237] Filling the variable entity content generated by the variable entity content generation model and the assignment results of the immutable entities into the corresponding question template to form a set of questions, and repeating this operation to generate question generation results for different knowledge and different question templates.

[0238] It should be understood that the above embodiments are only for illustration purposes and are not limited by this structure in the scope of the patent application.

[0239] It should be noted that the above serial numbers of the embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the term "including" or "comprising" or any other variant thereof in this article is intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, device, article or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, device, article or method including the element.

[0240] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium as described above (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0241] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. An intelligent question generation method for knowledge point analysis, characterized in that The method includes: S1: Obtain math problems, perform knowledge point annotation and entity annotation on the math problems, and cluster the math problems that examine the same knowledge based on the entity annotation results to obtain multiple math problem clusters that examine the same knowledge; Among them, the knowledge point annotation and entity annotation process for the nth group of math problems is: Annotate the knowledge examined by the nth group of math problems, where the knowledge that math problems can examine includes geometry, algebra, number theory, mathematical analysis, and probability and statistics, n ∈ [1, N]; Annotate the theorems and principles involved in the nth set of math problems, and form the knowledge point annotation result x of the nth set of math problems by combining the knowledge annotation result, theorem, and principle annotation result n : Among them: x n represents the knowledge point annotation result of the nth group of math problems, represents the knowledge examined by the nth group of math problems, represents the set of theorems and principles involved in the nth group of math problems; Identify the key elements in the nth group of math problems, where the types of key elements of the key elements include numbers, variables, geometric shapes, operators, and units. Mark the identified key elements and the corresponding types of the key elements to obtain the entity annotation result y of the nth group of math problems n : Among them: Represents the c-th key element in the n-th set of math problems, Represents a key element Type, C n Represents the total number of key elements in the n-th set of math problems, The key element types 1-5 are, in sequence, number, variable, geometric shape, operator, and unit; All the math problems with consistent knowledge under investigation form the math problem set of this knowledge, where the math problem set of the \(i\)-th kind of knowledge is \(\Omega\). i : Among them: It represents the s-th group of math problems for examining the i-th kind of knowledge, and Num(i) represents the total number of math problems for examining the i-th kind of knowledge; x 2 (s, i) represents the set of theorems and principles involved; Representation of the entity annotation result; represent the c-th key element of represent the key element of the key element type represent the number of key elements of Cluster the math problems that examine the same knowledge based on the entity annotation results to obtain multiple math problem clusters that examine the same knowledge, including: The clustering process for the math problems that examine the ith kind of knowledge is: S11: Calculate the set of math problems Ω i the distance between any two groups of math problems in, where the math problem and the distance between is: Among them: Represents a math problem and the distance between, a ∈ [1, Num(i)]; Represents a math problem and the distances between key elements in x 2 (s,i) ∩ x 2 (s,a) represents the set of theorems and principles x 2 (s,i) and x 2 the intersection between (s,a), x 2 (s, i) ∪ x 2 (s, a) represents the set of theorems and principles x 2 (s, i) and x 2 the union between (s, a) and x 2 (s, a) represents the set of theorems and principles involved; count(·) represents the number of theorems and principles in the set; representation entity annotation result of exp(·) represents the exponential function with the natural constant as the base; S12: Calculate the clustering center information weight of each group of math problems based on the distance between math problems, where the clustering center information weight of the math problems is Among them: For a math problem density; Denote the set of math problems as Ω i the number of math problems in that are less than a preset distance threshold from the math problem Denote the set of math problems as Ω i and the set of math problems in with a distance less than a preset distance threshold from the math problem Let e denote any math problem in the set of math problems ; Denote the math problem as the distance between Represents a math problem The clustering center coefficient of; Denote the set of mathematical problems as Ω i with a density greater than among the mathematical problems; Max_dis represents the preset maximum clustering center coefficient; S13: Select the math problems with the weight of clustering center information higher than the preset threshold as the math problem cluster centers, and merge and cluster the math problems that examine the knowledge of the i-th type and are non-math problem cluster centers with the nearest math problem cluster centers to form math problem clusters, where the set of math problem clusters that examine the knowledge of the i-th type is E i : Among them: Denote the \(k\)-th cluster of math questions for investigating the \(i\)-th type of knowledge, \(K\) i Denote the total number of clusters of math questions for investigating the \(i\)-th type of knowledge; S2: Use the association rule template mining method to extract templates from the math problem clusters to obtain the problem templates of the math problem clusters, annotate the entity variables in the problem templates to obtain the variable entities and immutable entities in the problem templates; Among them, the template extraction process of the mathematics problem cluster is as follows: S21: Tokenize and remove stop words from the math problem clusters in the math problems to obtain a sequence of phrases for the math problems. The sequence of phrases contains phrases with entity annotations as key elements and phrases that are not key elements. The math problem clusters The sequence of phrases for the m-th group of math problems in the math problem clusters is Among them: Denote the h-th phrase in the phrase sequence where h ∈ [1, H m , H m Denote the phrase sequence and the total number of phrases in it; Indicates the number of math problems in the math problem cluster; If then the phrase is a key element, and if then the phrase is not a key element; S22: Replace the phrases that are key elements in the phrase sequence with the key element types to obtain the phrase sequence after key element replacement; S23: Form a phrase set from multiple phrases in the phrase sequence, and calculate the support of each phrase set, where the minimum number of key element types in each phrase set is The upper limit of the number of phrases in the phrase set is The lower limit is is a preset phrase set threshold, where the support of phrase set Q is Support(Q): Q = (Q1, Q2,..., Q u ,..., Q U ); Among them: Q1, Q2,..., Q u ,..., Q U represent U phrases in the phrase set Q, where u ∈ [1, U], and U represents the total number of phrases in the phrase set Q; Sum(Q) represents the number of math problems in the math problem cluster that contain the phrases Q1, Q2,..., Q u ,..., Q U , and the order of appearance of the phrases is the same as that in the phrase set Q; S24: Use the phrase set with support greater than the preset support threshold as the question template for the math question cluster, and use the phrases in the question template as entity variables; ​ S3: Construct a variable entity content generation model, and use the variable entity content generation model to generate the variable entity content in the problem template and the assignment results of the immutable entities; S4: Fill the variable entity content and the assignment results of the immutable entities into the problem template to form problems, and obtain the problem generation results of different knowledge points.

2. The intelligent question generation method for knowledge point analysis according to claim 1, wherein The annotation of the entity variables in the problem template to obtain the variable entities and immutable entities in the problem template includes: Annotate the entity variables that are of the key element type in the problem template as immutable entities; Annotate the entity variables that are not of the key element type in the problem template as variable entities.

3. The intelligent question generation method for knowledge point analysis according to claim 1, wherein The S3 step includes: Construct a variable entity content generation model, where the variable entity content generation model takes the problem template as the input and the variable entity content and the assignment results of the immutable entities as the output. The variable entity content generation model includes an input layer, a context extraction layer, an iterative generation layer, and an output layer. The input layer is used to receive the problem template and the entity variable annotation results. The context extraction layer is used to extract the context immutable entities of the variable entities in the problem template and assign values to the context immutable entities. The iterative generation layer iteratively generates the variable entity content of the variable entities based on the context immutable entities to obtain the variable entity content of the variable entities in the problem template. The output layer outputs the assignment results of the immutable entities and the variable entity content.

4. The intelligent question generation method for knowledge point analysis according to claim 3, wherein The use of the variable entity content generation model to generate the variable entity content in the problem template and the assignment results of the immutable entities includes: S31: The input layer receives the problem template B and the entity variable annotation result β: B = (B1, B2,..., B v ,..., B V ); β＝(β1,β2,...,β v ,...,β V )； Among them: B v represents the v-th entity variable in question template B, where v ∈ [1, V] and V represents the total number of entity variables in question template B; β v is the annotation result of the entity variable B v . For β v = 1, it means that the entity variable B v is a variable entity. For β v = -1, it means that the entity variable B v is an immutable entity; S32: The context extraction layer extracts the context immutable entities of the variable entities in the question template B. The context immutable entities of the variable entities include the upper context immutable entities and the lower context immutable entities; For variable entity B v , then B v 's upper immutable entity is The lower immutable entity is Where: Denote B v the immediately preceding immutable entity of Denote B v the immediately following immutable entity of; S33: Assign values to context-immutable entities, where the above-context immutable entities and the below-context immutable entities are assigned values in the following way: Obtain the key element type, generate phrases that conform to this key element type, and perform assignment to obtain the immutable entity of the above text after assignment Obtain the key element type, generate phrases that conform to this key element type, and for perform assignment to obtain the immutable entity of the subsequent text after assignment S34: The iterative generation layer iteratively generates the variable entity content of the variable entity based on the context-immutable entity, and obtains the variable entity content of the variable entity in the question template, where the variable entity B v The iterative generation process of the variable entity content is as follows: S341: For and perform word vector representation to obtain a word vector sequence S342: Set the current iteration number of the word vector sequence corresponding to the variable entity content to t. The initial value of t is 0, and the maximum value is Max. Then, for variable entity B v the t-th iteration result of the word vector sequence corresponding to the variable entity content is which represent the upper-context word vector sequence and the lower-context word vector sequence obtained in the t-th iteration, respectively; S343: Perform semantic iterative processing with random perturbations added to the word vector sequence corresponding to the variable entity content. The semantic iterative processing formula is: Where: T represents transpose; W represents the semantic convolution matrix; * represents the convolution operator; δ(·) represents the ReLU activation function; σ represents the random noise sequence; I represents the identity matrix; S344: Let t = t + 1, return to step S343 until the maximum number of iterations is reached, and use the word vector sequence obtained by semantic iteration processing at this time Convert it into the form of a phrase sequence, and use the word vector sequence above The last phrase of the corresponding phrase sequence as the variable entity B v as the variable entity content of S35: Extract the assignment results of all immutable entities in step S33 and the variable entity content of all variable entities in step S34. The output layer outputs the assignment results of the immutable entities and the variable entity content as a set of question template content to be filled.

5. The intelligent question generation method for knowledge point analysis according to claim 4, wherein The S4 step includes: Generate the variable entity content generated by the variable entity content generation model and the assignment results of the immutable entities, fill them into the corresponding question templates to form a set of questions, and repeat this operation to generate question generation results under different knowledge and different question templates.

6. The intelligent question generation method for knowledge point analysis according to claim 2, wherein The calculation process is as follows: Initialize C[b s [0]=0, C[0][b a =0, where represents the number of key elements of the math problem ; Perform recursion on C[b s [b a , where C[b s [b a represents the distance between the first b key elements of math problem s and the first b key elements of math problem a The recurrence formula is: Where: Represents a math problem of the b s th key element type of the key element; Represents the key element type of the b -th key element of the math problem a ; max(C[b s -1][b a ,C[b s [b a -1]) represents selecting the maximum value from C[b s -1][b a , C[b s [b a -1]; Calculate the math problem based on the recursively obtained distance and the distances between the key elements in Where: Represents a math problem and the distances between key elements in

Images