Disclosure of Invention
The technical problem to be solved by the invention is to provide a test question generation method and a test question generation system based on a knowledge graph and a prediction model, which can realize more intelligent and automatic test question generation, and the generated test questions can give users a certain difficulty challenge to improve the learning effect.
In order to solve the technical problems, the first aspect of the invention discloses a test question generation method based on a knowledge graph and a prediction model, which comprises the following steps:
Acquiring examination requirements and user parameters of a target user;
according to the examination requirement, determining a plurality of candidate test question sets from a candidate test question database based on an identification matching algorithm;
predicting user level parameters of the target user based on a trained level prediction neural network model according to the examination requirements and the user parameters;
according to the examination requirement, determining a plurality of candidate knowledge points corresponding to the target user in a pre-built examination-level-knowledge point associated knowledge map;
screening a plurality of target knowledge points with knowledge point difficulty higher than the user level parameters from the plurality of candidate knowledge points according to the user level parameters;
and determining the examination question set corresponding to the target user according to the target knowledge point and the plurality of candidate question sets.
As an optional implementation manner, in the first aspect of the present invention, the test requirement includes at least one of a name of a ready-to-take test, a name of a completed test, a level of a test, a level of education corresponding to the test, a level of school year corresponding to the test, and a region corresponding to the test; and determining a plurality of candidate test question sets from a candidate test question database based on an identification matching algorithm according to the test requirement, wherein the method comprises the following steps:
Performing vector conversion on the examination demands based on a trained vector conversion model to obtain demand vector representations corresponding to the examination demands;
for any test question set in the candidate test question database, acquiring a test question set identifier corresponding to the test question set;
inputting the test question set identification into the vector conversion model to obtain an identification vector representation corresponding to the test question set;
calculating a vector distance between the demand vector representation and the identification vector representation;
calculating the occurrence times of each test question in the test question sets in other test question sets; the occurrence number comprises the number of the same or judged similar test questions in other test question sets;
calculating the repeatability adjustment weight corresponding to the frequency average value of the occurrence times of all the test questions in the test question set; the repetition degree adjustment weight is inversely proportional to the number average;
calculating the product of the vector distance and the repeatability adjustment weight to obtain a matching parameter corresponding to the test question set;
sequencing all the test question sets according to the matching parameters from high to low to obtain a test question set sequence;
and determining a preset number of test question sets with the matching parameters larger than a preset parameter threshold value in the test question set sequence as candidate test question sets so as to obtain a plurality of candidate test question sets.
As an optional implementation manner, in the first aspect of the present invention, the user parameter includes at least one of a user age, a user gender, a user physical parameter, a user education level, and a user history error question record;
and predicting user level parameters of the target user based on the trained level prediction neural network model according to the examination requirements and the user parameters, including:
inputting the examination requirement to a trained horizontal prediction neural network model to obtain a first horizontal parameter corresponding to the target user and a corresponding first prediction probability value;
inputting the user parameters into the horizontal prediction neural network model to obtain second horizontal parameters corresponding to the target user and second prediction probability values corresponding to the target user;
calculating a parameter difference between the first level parameter and the second level parameter, and a probability difference between the first predicted probability value and the second predicted probability value;
calculating the ratio of the parameter difference value to the probability difference value, and judging whether the ratio is larger than a preset ratio threshold value or not to obtain a first judging result;
if the first judgment result is yes, determining the second level parameter as a user level parameter corresponding to the target user, and if the first judgment result is no, judging whether the probability difference is larger than a preset difference threshold value or not, so as to obtain a second judgment result;
And if the second judgment result is yes, determining a horizontal parameter corresponding to a probability value with a higher numerical value in the first predicted probability value and the second predicted probability value as the user horizontal parameter, and if the second judgment result is no, determining an average value of the first horizontal parameter and the second horizontal parameter as the user horizontal parameter.
In an optional implementation manner, in a first aspect of the present invention, the determining, according to the test requirement, a plurality of candidate knowledge points corresponding to the target user in a pre-built knowledge graph of test-level-knowledge point association includes:
based on a preset knowledge framework and a principal component analysis algorithm, establishing a knowledge graph model of examination-level-knowledge point association according to a plurality of knowledge point data, corresponding examination application records and examinee parameter records; the knowledge graph model comprises a plurality of knowledge points forming different learning chains; each user level feature combination corresponds to a plurality of knowledge points; each examination parameter feature combination corresponds to a plurality of knowledge points;
calculating first similarity between the examination requirement and any examination parameter feature combination, and determining all knowledge points corresponding to all examination parameter feature combinations with the first similarity larger than a preset first similarity threshold as a plurality of candidate knowledge points corresponding to the target user.
In a first aspect of the present invention, the establishing a knowledge graph model of test-level-knowledge point association according to a plurality of knowledge point data and corresponding test application records and test taker parameter records based on a preset knowledge frame and a principal component analysis algorithm includes:
acquiring a plurality of knowledge point data and examination application records and test taker parameter records corresponding to each knowledge point data;
forming node-stage relations corresponding to a plurality of knowledge point data according to a preset knowledge frame to obtain a basic knowledge graph model;
according to a plurality of item factors and corresponding item achievements of a plurality of historical examination items in examination application records corresponding to all the knowledge point data, screening out a plurality of factor characteristic combinations corresponding to the plurality of item factors based on a principal component analysis algorithm so as to obtain a plurality of examination parameter characteristic combinations; the project factors comprise at least one of examination names, examination levels, education levels corresponding to examination, school year levels corresponding to examination and areas corresponding to examination;
based on a preset examination level grade grading rule, determining a plurality of history examination items corresponding to each examination level grade;
Screening out a plurality of user level feature combinations corresponding to the plurality of examinee factors based on a principal component analysis algorithm according to the plurality of examinee factors and the corresponding examination results in the examinee parameter records of all the history examination items so as to obtain a plurality of user level feature combinations corresponding to the examination level grade; the examinee parameters comprise at least one of examinee age, examinee gender, examinee physical parameters and examinee education degree.
In a first aspect of the present invention, the screening, according to the user level parameter, a plurality of target knowledge points with knowledge point difficulty higher than the user level parameter from the plurality of candidate knowledge points includes:
determining the test level grade closest to the user level parameter as a first test level grade;
determining the examination level grade higher than the first examination level grade by a preset level difference as a target examination level grade;
and calculating a second similarity between each user level feature combination corresponding to the target examination level and the user parameter, and determining all knowledge points corresponding to all user level feature combinations with the second similarity being larger than a preset similarity threshold value as a plurality of target knowledge points corresponding to the target user.
In an optional implementation manner, in a first aspect of the present invention, the determining, according to the target knowledge point and the plurality of candidate test question sets, a test question set corresponding to the target user includes:
obtaining corresponding vector representations from any two target knowledge points to the vector conversion model, and calculating vector similarity to obtain a third similarity between the two target knowledge points;
calculating the number of other target knowledge points, corresponding to each target knowledge point, of which the third similarity is larger than a preset third similarity threshold, and screening a plurality of low-repetition-degree knowledge points from all the target knowledge points according to the number;
for any test question and any low-repetition-degree knowledge point in any candidate test question set, inputting the test question and the low-repetition-degree knowledge point into the vector conversion model to obtain corresponding vector representation and calculating vector similarity so as to obtain fourth similarity between the test question and the low-repetition-degree knowledge point;
screening out all the test questions with the fourth similarity higher than a preset fourth similarity threshold value corresponding to the low-repetition degree knowledge points for each low-repetition degree knowledge point to obtain a plurality of target test questions corresponding to the low-repetition degree knowledge points;
And determining all the target test questions corresponding to all the low-repetition-degree knowledge points as test question sets corresponding to the target users.
The invention discloses a test question generation system based on a knowledge graph and a prediction model, which comprises the following components:
the acquisition module is used for acquiring examination requirements and user parameters of the target user;
the matching module is used for determining a plurality of candidate test question sets from a candidate test question database based on an identification matching algorithm according to the test requirements;
the prediction module is used for predicting the user level parameters of the target user based on a trained level prediction neural network model according to the examination requirements and the user parameters;
the first determining module is used for determining a plurality of candidate knowledge points corresponding to the target user in a pre-built knowledge map of examination-level-knowledge point association according to the examination requirement;
the screening module is used for screening a plurality of target knowledge points with knowledge point difficulty higher than the user level parameters from the plurality of candidate knowledge points according to the user level parameters;
and the second determining module is used for determining the examination question set corresponding to the target user according to the target knowledge point and the plurality of candidate test question sets.
As an optional implementation manner, in the second aspect of the present invention, the examination requirement includes at least one of a name of an examination to be attended, a name of an already completed examination, a level of an examination, a level of education corresponding to the examination, a level of school year corresponding to the examination, and a region corresponding to the examination; and the matching module determines a specific mode of a plurality of candidate test question sets from a candidate test question database based on an identification matching algorithm according to the examination requirement, and the specific mode comprises the following steps:
performing vector conversion on the examination demands based on a trained vector conversion model to obtain demand vector representations corresponding to the examination demands;
for any test question set in the candidate test question database, acquiring a test question set identifier corresponding to the test question set;
inputting the test question set identification into the vector conversion model to obtain an identification vector representation corresponding to the test question set;
calculating a vector distance between the demand vector representation and the identification vector representation;
calculating the occurrence times of each test question in the test question sets in other test question sets; the occurrence number comprises the number of the same or judged similar test questions in other test question sets;
Calculating the repeatability adjustment weight corresponding to the frequency average value of the occurrence times of all the test questions in the test question set; the repetition degree adjustment weight is inversely proportional to the number average;
calculating the product of the vector distance and the repeatability adjustment weight to obtain a matching parameter corresponding to the test question set;
sequencing all the test question sets according to the matching parameters from high to low to obtain a test question set sequence;
and determining a preset number of test question sets with the matching parameters larger than a preset parameter threshold value in the test question set sequence as candidate test question sets so as to obtain a plurality of candidate test question sets.
As an optional implementation manner, in the second aspect of the present invention, the user parameter includes at least one of a user age, a user gender, a user physical parameter, a user education level, and a user history error question record;
and the prediction module predicts a specific mode of the user level parameter of the target user based on the trained level prediction neural network model according to the examination requirement and the user parameter, and comprises the following steps:
inputting the examination requirement to a trained horizontal prediction neural network model to obtain a first horizontal parameter corresponding to the target user and a corresponding first prediction probability value;
Inputting the user parameters into the horizontal prediction neural network model to obtain second horizontal parameters corresponding to the target user and second prediction probability values corresponding to the target user;
calculating a parameter difference between the first level parameter and the second level parameter, and a probability difference between the first predicted probability value and the second predicted probability value;
calculating the ratio of the parameter difference value to the probability difference value, and judging whether the ratio is larger than a preset ratio threshold value or not to obtain a first judging result;
if the first judgment result is yes, determining the second level parameter as a user level parameter corresponding to the target user, and if the first judgment result is no, judging whether the probability difference is larger than a preset difference threshold value or not, so as to obtain a second judgment result;
and if the second judgment result is yes, determining a horizontal parameter corresponding to a probability value with a higher numerical value in the first predicted probability value and the second predicted probability value as the user horizontal parameter, and if the second judgment result is no, determining an average value of the first horizontal parameter and the second horizontal parameter as the user horizontal parameter.
In a second aspect of the present invention, the specific manner of determining, by the first determining module, a plurality of candidate knowledge points corresponding to the target user in a pre-built knowledge graph of examination-level-knowledge point association according to the examination requirement includes:
based on a preset knowledge framework and a principal component analysis algorithm, establishing a knowledge graph model of examination-level-knowledge point association according to a plurality of knowledge point data, corresponding examination application records and examinee parameter records; the knowledge graph model comprises a plurality of knowledge points forming different learning chains; each user level feature combination corresponds to a plurality of knowledge points; each examination parameter feature combination corresponds to a plurality of knowledge points;
calculating first similarity between the examination requirement and any examination parameter feature combination, and determining all knowledge points corresponding to all examination parameter feature combinations with the first similarity larger than a preset first similarity threshold as a plurality of candidate knowledge points corresponding to the target user.
In a second aspect of the present invention, the first determining module establishes a knowledge graph model of test-level-knowledge point association according to a plurality of knowledge point data and corresponding test application records and test taker parameter records based on a preset knowledge frame and a principal component analysis algorithm, and the specific manner of establishing the knowledge graph model of test-level-knowledge point association includes:
Acquiring a plurality of knowledge point data and examination application records and test taker parameter records corresponding to each knowledge point data;
forming node-stage relations corresponding to a plurality of knowledge point data according to a preset knowledge frame to obtain a basic knowledge graph model;
according to a plurality of item factors and corresponding item achievements of a plurality of historical examination items in examination application records corresponding to all the knowledge point data, screening out a plurality of factor characteristic combinations corresponding to the plurality of item factors based on a principal component analysis algorithm so as to obtain a plurality of examination parameter characteristic combinations; the project factors comprise at least one of examination names, examination levels, education levels corresponding to examination, school year levels corresponding to examination and areas corresponding to examination;
based on a preset examination level grade grading rule, determining a plurality of history examination items corresponding to each examination level grade;
screening out a plurality of user level feature combinations corresponding to the plurality of examinee factors based on a principal component analysis algorithm according to the plurality of examinee factors and the corresponding examination results in the examinee parameter records of all the history examination items so as to obtain a plurality of user level feature combinations corresponding to the examination level grade; the examinee parameters comprise at least one of examinee age, examinee gender, examinee physical parameters and examinee education degree.
In a second aspect of the present invention, the specific manner of the screening module for screening, according to the user level parameter, a plurality of target knowledge points with knowledge point difficulty higher than the user level parameter from the plurality of candidate knowledge points includes:
determining the test level grade closest to the user level parameter as a first test level grade;
determining the examination level grade higher than the first examination level grade by a preset level difference as a target examination level grade;
and calculating a second similarity between each user level feature combination corresponding to the target examination level and the user parameter, and determining all knowledge points corresponding to all user level feature combinations with the second similarity being larger than a preset similarity threshold value as a plurality of target knowledge points corresponding to the target user.
In a second aspect of the present invention, the second determining module determines, according to the target knowledge point and the plurality of candidate test question sets, a specific manner of the test question set corresponding to the target user, including:
obtaining corresponding vector representations from any two target knowledge points to the vector conversion model, and calculating vector similarity to obtain a third similarity between the two target knowledge points;
Calculating the number of other target knowledge points, corresponding to each target knowledge point, of which the third similarity is larger than a preset third similarity threshold, and screening a plurality of low-repetition-degree knowledge points from all the target knowledge points according to the number;
for any test question and any low-repetition-degree knowledge point in any candidate test question set, inputting the test question and the low-repetition-degree knowledge point into the vector conversion model to obtain corresponding vector representation and calculating vector similarity so as to obtain fourth similarity between the test question and the low-repetition-degree knowledge point;
screening out all the test questions with the fourth similarity higher than a preset fourth similarity threshold value corresponding to the low-repetition degree knowledge points for each low-repetition degree knowledge point to obtain a plurality of target test questions corresponding to the low-repetition degree knowledge points;
and determining all the target test questions corresponding to all the low-repetition-degree knowledge points as test question sets corresponding to the target users.
The third aspect of the invention discloses another test question generation system based on a knowledge graph and a prediction model, which comprises:
a memory storing executable program code;
A processor coupled to the memory;
the processor invokes the executable program codes stored in the memory to execute part or all of the steps in the test question generation method based on the knowledge graph and the prediction model disclosed in the first aspect of the invention.
A fourth aspect of the present invention discloses a computer storage medium storing computer instructions for executing part or all of the steps of the method for generating questions based on knowledge graph and predictive model disclosed in the first aspect of the present invention when the computer instructions are called.
Compared with the prior art, the invention has the following beneficial effects:
according to the invention, a neural network algorithm and a knowledge graph model can be combined, and a test question set with higher difficulty is determined from a preset test question library according to the examination requirement of a user and the user parameters, so that more intelligent and automatic test question generation can be realized, and the generated test questions can give the user a certain difficulty challenge to improve the learning effect.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The terms first, second and the like in the description and in the claims and in the above-described figures are used for distinguishing between different objects and not necessarily for describing a sequential or chronological order. Furthermore, the terms "comprise" and "have," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those listed steps or elements but may include other steps or elements not listed or inherent to such process, method, article, or apparatus.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the invention. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.
The invention discloses a test question generation method and a system based on a knowledge graph and a prediction model, which can combine a neural network algorithm and the knowledge graph model, determine a test question set with higher difficulty from a preset test question library according to the examination requirement of a user and the user parameters, so that more intelligent and automatic test question generation can be realized, and the generated test questions can give the user a certain difficulty challenge to improve the learning effect. The following will describe in detail.
Referring to fig. 1, fig. 1 is a schematic flow chart of a test question generating method based on a knowledge graph and a prediction model according to an embodiment of the present invention. The method described in fig. 1 may be applied to a corresponding data processing device, a data processing terminal, and a data processing server, where the server may be a local server or a cloud server, and the embodiment of the present invention is not limited to the method shown in fig. 1, and the method for generating a test question based on a knowledge graph and a prediction model may include the following operations:
101. And acquiring examination requirements and user parameters of the target user.
Optionally, the target user may be student users or self-learning adult users with different age groups, and the examination requirement and user parameters may be obtained by analyzing historical user data or operation records, or may be input by the user.
102. And determining a plurality of candidate test question sets from the candidate test question database based on the identification matching algorithm according to the test requirement.
Optionally, the test requirement includes at least one of a name of the test to be attended, a name of the test already completed, a level of education corresponding to the test, a level of school year corresponding to the test, and a region corresponding to the test, and the identification matching algorithm may perform matching calculation directly with the test requirement by using an identification corresponding to the candidate test question set, and determine the test question set with higher matching degree as the candidate test question set.
103. And predicting the user level parameters of the target user based on the trained level prediction neural network model according to the examination requirements and the user parameters.
Optionally, the user parameter includes at least one of a user age, a user gender, a user physical parameter, a user education level, a user history error question record. Alternatively, the examination requirements and the user parameters can be directly input into the trained horizontal prediction neural network model to predict the user horizontal parameters of the target user.
104. And determining a plurality of candidate knowledge points corresponding to the target user in the pre-built knowledge map of examination-level-knowledge point association according to examination requirements.
105. And screening a plurality of target knowledge points with knowledge point difficulty higher than the user level parameters from the plurality of candidate knowledge points according to the user level parameters.
106. And determining the examination question set corresponding to the target user according to the target knowledge point and the plurality of candidate question sets.
Therefore, the method described by the embodiment of the invention can be combined with the neural network algorithm and the knowledge graph model, and the test question set with higher difficulty can be determined in the preset test question library according to the examination requirement of the user and the user parameters, so that more intelligent and automatic test question generation can be realized, and the generated test questions can give the user a certain difficulty challenge to improve the learning effect.
As an optional embodiment, in the step, according to the examination requirement, determining, based on the identification matching algorithm, a plurality of candidate test question sets from a candidate test question database includes:
performing vector conversion on the examination requirement based on the trained vector conversion model to obtain a requirement vector representation corresponding to the examination requirement;
For any test question set in the candidate test question database, acquiring a test question set identifier corresponding to the test question set;
inputting the test question set identification into a vector conversion model to obtain an identification vector representation corresponding to the test question set;
calculating a vector distance between the demand vector representation and the identification vector representation;
calculating the occurrence times of each test question in the test question sets in other test question sets; the number of occurrences includes the number of identical or judged similar questions for each question in the other question sets;
calculating the repeatability adjustment weight corresponding to the frequency average value of the occurrence frequency of all the test questions in the test question set; the repetition degree adjustment weight is inversely proportional to the number average;
calculating the product of the vector distance and the repeatability adjustment weight to obtain a matching parameter corresponding to the test question set;
sequencing all the test question sets according to the matching parameters from high to low to obtain a test question set sequence;
and determining a plurality of test question sets which are preset in the test question set sequence and have the matching parameters larger than a preset parameter threshold value as candidate test question sets so as to obtain a plurality of candidate test question sets.
Through the embodiment, the matching parameters of different test question sets and examination demands can be calculated through more accurate algorithm steps and weight adjustment, so that more matched candidate test question sets are obtained through screening, the subsequently generated test questions more accord with the examination demands of users, and a better learning and examination effect is achieved.
As an optional embodiment, in the step, according to the test requirement and the user parameter, the predicting the user level parameter of the target user based on the trained level prediction neural network model includes:
inputting examination requirements into a trained horizontal prediction neural network model to obtain a first horizontal parameter corresponding to a target user and a corresponding first prediction probability value;
inputting the user parameters into a horizontal prediction neural network model to obtain a second horizontal parameter corresponding to the target user and a second prediction probability value corresponding to the target user;
calculating a parameter difference between the first horizontal parameter and the second horizontal parameter, and a probability difference between the first predicted probability value and the second predicted probability value;
calculating the ratio of the parameter difference value and the probability difference value, and judging whether the ratio is larger than a preset ratio threshold value or not to obtain a first judging result;
if the first judgment result is yes, determining the second level parameter as a user level parameter corresponding to the target user, and if the first judgment result is no, judging whether the probability difference is larger than a preset difference threshold value, so as to obtain a second judgment result;
if the second judgment result is yes, determining a horizontal parameter corresponding to a probability value with a higher numerical value in the first predicted probability value and the second predicted probability value as a user horizontal parameter, and if the second judgment result is no, determining an average value of the first horizontal parameter and the second horizontal parameter as the user horizontal parameter.
Alternatively, the horizontal prediction neural network model may be trained by a training data set including training test requirement parameters and training user parameters and corresponding user level class labels, and may be a model of a CNN structure, RNN structure or LTSM structure, or a random forest algorithm model including multiple models. Optionally, the training data set, when labeling the user level, carries out the user grading according to the preset grading rule of the examination level, and the grading can be completed by professional education personnel.
Through the embodiment, the level parameters corresponding to the user can be determined through various rules such as neural network models, difference judgment of prediction possibility and the like, when the level difference predicted by the two parameters is too large and the probability is also extremely large, namely, when the predicted result is in an unbalanced state, the result of the user parameter is directly used as the standard, when the difference of the level and the difference of the probability are not very balanced, whether the difference of the two probabilities is large or not is considered, averaging is carried out under the condition that the difference of the level difference and the probability is not large, and when the difference of the two probabilities is large, the predicted result of high probability is used as the final result, so that the level of the user can be accurately predicted, the subsequently generated test question can be accurately higher than the level of the user, and a better learning test effect is achieved.
As an optional embodiment, in the step, according to the examination requirement, determining, in a pre-built knowledge graph of examination-level-knowledge point association, a plurality of candidate knowledge points corresponding to the target user includes:
based on a preset knowledge framework and a principal component analysis algorithm, establishing a knowledge graph model of examination-level-knowledge point association according to a plurality of knowledge point data, corresponding examination application records and examinee parameter records; the knowledge graph model comprises a plurality of knowledge points forming different learning chains; each user level feature combination corresponds to a plurality of knowledge points; each examination parameter feature combination corresponds to a plurality of knowledge points;
calculating first similarity between the examination requirement and any examination parameter feature combination, and determining all knowledge points corresponding to all examination parameter feature combinations with the first similarity larger than a preset first similarity threshold as a plurality of candidate knowledge points corresponding to the target user.
Through the embodiment, all knowledge points corresponding to all examination parameter feature combinations with the first similarity larger than the preset first similarity threshold can be determined as a plurality of candidate knowledge points corresponding to the target user by calculating the first similarity between the examination requirement and any examination parameter feature combination, so that the corresponding knowledge points can be accurately determined from the knowledge graph, and the follow-up examination questions generated according to the knowledge point screening can be accurately aimed at the examination requirement, thereby achieving better learning and examination effects.
As an optional embodiment, in the step, based on a preset knowledge frame and a principal component analysis algorithm, a knowledge graph model of test-level-knowledge point association is established according to a plurality of knowledge point data and corresponding test application records and test taker parameter records, and the method includes:
acquiring a plurality of knowledge point data and examination application records and test taker parameter records corresponding to each knowledge point data;
forming node-stage relations corresponding to a plurality of knowledge point data according to a preset knowledge frame to obtain a basic knowledge graph model;
screening out a plurality of factor characteristic combinations corresponding to a plurality of item factors based on a principal component analysis algorithm according to a plurality of item factors and corresponding item achievements of a plurality of historical examination items in examination application records corresponding to all knowledge point data so as to obtain a plurality of examination parameter characteristic combinations; the project factors comprise at least one of examination names, examination levels, education levels corresponding to examination, school year levels corresponding to examination and areas corresponding to examination;
based on a preset examination level grade grading rule, determining a plurality of history examination items corresponding to each examination level grade;
Screening a plurality of user level feature combinations corresponding to the plurality of examinee factors based on a principal component analysis algorithm according to the plurality of examinee factors and the corresponding examination results in the examinee parameter records of all the history examination projects so as to obtain a plurality of user level feature combinations corresponding to the examination level grade; the examinee parameters comprise at least one of examinee age, examinee gender, examinee physical parameters and examinee education degree.
Through the embodiment, based on the preset knowledge framework and the principal component analysis algorithm, the knowledge graph model of examination-level-knowledge point association can be established according to the plurality of knowledge point data, the corresponding examination application records and the examinee parameter records, so that the knowledge graph model can establish a good and sufficient data basis for subsequent knowledge point screening, and the matching degree of the finally generated test questions is improved.
As an optional embodiment, in the step, according to the user level parameter, a plurality of target knowledge points with knowledge point difficulty higher than the user level parameter are selected from a plurality of candidate knowledge points, including:
determining the test level grade closest to the user level parameter as a first test level grade;
Determining the examination level grade higher than the first examination level grade by a preset grade difference as a target examination level grade;
and calculating a second similarity between each user level characteristic combination corresponding to the target examination level and the user parameters, and determining all knowledge points corresponding to all user level characteristic combinations with the second similarity larger than a preset similarity threshold as a plurality of target knowledge points corresponding to the target user.
Through the embodiment, the second similarity between each user level feature combination corresponding to the target examination level and the user parameters can be calculated, all knowledge points corresponding to all user level feature combinations with the second similarity being larger than the preset two similarity threshold value are determined as a plurality of target knowledge points corresponding to the target user, so that the corresponding knowledge points higher than the user level are accurately determined from the knowledge map, and the follow-up test questions generated according to the knowledge point screening can give certain challenges to the user, so that a better learning and testing effect is achieved.
As an optional embodiment, in the step, determining the test question set corresponding to the target user according to the target knowledge point and the plurality of candidate test question sets includes:
Obtaining corresponding vector representations from any two target knowledge points to a vector conversion model and calculating vector similarity to obtain a third similarity between the two target knowledge points;
calculating the number of other target knowledge points, corresponding to each target knowledge point, of which the third similarity is larger than a preset third similarity threshold, and screening a plurality of low-repetition-degree knowledge points from all the target knowledge points according to the number;
for any test question and any low-repetition degree knowledge point in any candidate test question set, inputting the test question and the low-repetition degree knowledge point into a vector conversion model to obtain corresponding vector representation and calculate vector similarity so as to obtain fourth similarity between the test question and the low-repetition degree knowledge point;
screening out all test questions with the fourth similarity higher than a preset fourth similarity threshold value corresponding to the low-repetition degree knowledge points for each low-repetition degree knowledge point to obtain a plurality of target test questions corresponding to the low-repetition degree knowledge points;
and determining all target test questions corresponding to all the low-repetition knowledge points as an examination test question set corresponding to the target user.
Through the embodiment, the knowledge points with low repetition degree can be screened out firstly to avoid excessive consumption of computational resources in the subsequent test question matching step or excessive repetition of the subsequently generated test questions, and then the test questions which are more similar to the knowledge points are screened out from the candidate test question library, so that the test questions generated by screening according to the knowledge points can give a certain challenge to a user, and a better learning test effect is achieved.
In a second embodiment, referring to fig. 2, fig. 2 is a schematic structural diagram of a test question generating system based on a knowledge graph and a prediction model according to an embodiment of the present invention. The system described in fig. 2 may be applied to a corresponding data processing device, a data processing terminal, and a data processing server, where the server may be a local server or a cloud server, and embodiments of the present invention are not limited. As shown in fig. 2, the system may include:
an acquisition module 201, configured to acquire an examination requirement and a user parameter of a target user;
the matching module 202 is configured to determine, according to an examination requirement, a plurality of candidate test question sets from a candidate test question database based on an identification matching algorithm;
the prediction module 203 is configured to predict a user level parameter of a target user based on the trained level prediction neural network model according to the test requirement and the user parameter;
the first determining module 204 is configured to determine, according to an examination requirement, a plurality of candidate knowledge points corresponding to the target user in a pre-built knowledge graph associated with examination-level-knowledge points;
the screening module 205 is configured to screen, according to the user level parameter, a plurality of target knowledge points with knowledge point difficulty higher than the user level parameter from a plurality of candidate knowledge points;
The second determining module 206 is configured to determine an examination question set corresponding to the target user according to the target knowledge point and the plurality of candidate test question sets.
As an alternative embodiment, the examination requirement includes at least one of a name of an examination to be attended, a name of an already completed examination, a level of examination, an education level corresponding to the examination, a level of school year corresponding to the examination, and a region corresponding to the examination; and, the matching module 202 determines a specific mode of a plurality of candidate test question sets from the candidate test question database based on the identification matching algorithm according to the test requirement, including:
performing vector conversion on the examination requirement based on the trained vector conversion model to obtain a requirement vector representation corresponding to the examination requirement;
for any test question set in the candidate test question database, acquiring a test question set identifier corresponding to the test question set;
inputting the test question set identification into a vector conversion model to obtain an identification vector representation corresponding to the test question set;
calculating a vector distance between the demand vector representation and the identification vector representation;
calculating the occurrence times of each test question in the test question sets in other test question sets; the number of occurrences includes the number of identical or judged similar questions for each question in the other question sets;
Calculating the repeatability adjustment weight corresponding to the frequency average value of the occurrence frequency of all the test questions in the test question set; the repetition degree adjustment weight is inversely proportional to the number average;
calculating the product of the vector distance and the repeatability adjustment weight to obtain a matching parameter corresponding to the test question set;
sequencing all the test question sets according to the matching parameters from high to low to obtain a test question set sequence;
and determining a plurality of test question sets which are preset in the test question set sequence and have the matching parameters larger than a preset parameter threshold value as candidate test question sets so as to obtain a plurality of candidate test question sets.
As an alternative embodiment, the user parameters include at least one of user age, user gender, user physical parameters, user education level, user history error questions;
and, the prediction module 203 predicts a specific manner of the user level parameter of the target user based on the trained level prediction neural network model according to the examination requirement and the user parameter, including:
inputting examination requirements into a trained horizontal prediction neural network model to obtain a first horizontal parameter corresponding to a target user and a corresponding first prediction probability value;
inputting the user parameters into a horizontal prediction neural network model to obtain a second horizontal parameter corresponding to the target user and a second prediction probability value corresponding to the target user;
Calculating a parameter difference between the first horizontal parameter and the second horizontal parameter, and a probability difference between the first predicted probability value and the second predicted probability value;
calculating the ratio of the parameter difference value and the probability difference value, and judging whether the ratio is larger than a preset ratio threshold value or not to obtain a first judging result;
if the first judgment result is yes, determining the second level parameter as a user level parameter corresponding to the target user, and if the first judgment result is no, judging whether the probability difference is larger than a preset difference threshold value, so as to obtain a second judgment result;
if the second judgment result is yes, determining a horizontal parameter corresponding to a probability value with a higher numerical value in the first predicted probability value and the second predicted probability value as a user horizontal parameter, and if the second judgment result is no, determining an average value of the first horizontal parameter and the second horizontal parameter as the user horizontal parameter.
As an optional embodiment, the specific manner of determining, by the first determining module 204, a plurality of candidate knowledge points corresponding to the target user in the pre-built knowledge graph of examination-level-knowledge point association according to the examination requirement includes:
based on a preset knowledge framework and a principal component analysis algorithm, establishing a knowledge graph model of examination-level-knowledge point association according to a plurality of knowledge point data, corresponding examination application records and examinee parameter records; the knowledge graph model comprises a plurality of knowledge points forming different learning chains; each user level feature combination corresponds to a plurality of knowledge points; each examination parameter feature combination corresponds to a plurality of knowledge points;
Calculating first similarity between the examination requirement and any examination parameter feature combination, and determining all knowledge points corresponding to all examination parameter feature combinations with the first similarity larger than a preset first similarity threshold as a plurality of candidate knowledge points corresponding to the target user.
As an alternative embodiment, the first determining module 204 establishes a specific manner of the knowledge graph model of the test-level-knowledge point association according to the plurality of knowledge point data and the corresponding test application records and the test taker parameter records based on the preset knowledge frame and the principal component analysis algorithm, and includes:
acquiring a plurality of knowledge point data and examination application records and test taker parameter records corresponding to each knowledge point data;
forming node-stage relations corresponding to a plurality of knowledge point data according to a preset knowledge frame to obtain a basic knowledge graph model;
screening out a plurality of factor characteristic combinations corresponding to a plurality of item factors based on a principal component analysis algorithm according to a plurality of item factors and corresponding item achievements of a plurality of historical examination items in examination application records corresponding to all knowledge point data so as to obtain a plurality of examination parameter characteristic combinations; the project factors comprise at least one of examination names, examination levels, education levels corresponding to examination, school year levels corresponding to examination and areas corresponding to examination;
Based on a preset examination level grade grading rule, determining a plurality of history examination items corresponding to each examination level grade;
screening a plurality of user level feature combinations corresponding to the plurality of examinee factors based on a principal component analysis algorithm according to the plurality of examinee factors and the corresponding examination results in the examinee parameter records of all the history examination projects so as to obtain a plurality of user level feature combinations corresponding to the examination level grade; the examinee parameters comprise at least one of examinee age, examinee gender, examinee physical parameters and examinee education degree.
As an alternative embodiment, the specific manner of screening, by the screening module 205, a plurality of target knowledge points with knowledge point difficulty higher than the user level parameter from a plurality of candidate knowledge points according to the user level parameter includes:
determining the test level grade closest to the user level parameter as a first test level grade;
determining the examination level grade higher than the first examination level grade by a preset grade difference as a target examination level grade;
and calculating a second similarity between each user level characteristic combination corresponding to the target examination level and the user parameters, and determining all knowledge points corresponding to all user level characteristic combinations with the second similarity larger than a preset similarity threshold as a plurality of target knowledge points corresponding to the target user.
As an optional embodiment, the second determining module 206 determines, according to the target knowledge point and the plurality of candidate test question sets, a specific manner of the test question set corresponding to the target user, including:
obtaining corresponding vector representations from any two target knowledge points to a vector conversion model and calculating vector similarity to obtain a third similarity between the two target knowledge points;
calculating the number of other target knowledge points, corresponding to each target knowledge point, of which the third similarity is larger than a preset third similarity threshold, and screening a plurality of low-repetition-degree knowledge points from all the target knowledge points according to the number;
for any test question and any low-repetition degree knowledge point in any candidate test question set, inputting the test question and the low-repetition degree knowledge point into a vector conversion model to obtain corresponding vector representation and calculate vector similarity so as to obtain fourth similarity between the test question and the low-repetition degree knowledge point;
screening out all test questions with the fourth similarity higher than a preset fourth similarity threshold value corresponding to the low-repetition degree knowledge points for each low-repetition degree knowledge point to obtain a plurality of target test questions corresponding to the low-repetition degree knowledge points;
And determining all target test questions corresponding to all the low-repetition knowledge points as an examination test question set corresponding to the target user.
In particular, the technical details or technical effects of the foregoing embodiments or modules may be referred to the description in the first embodiment, and are not repeated herein.
Referring to fig. 3, fig. 3 is a schematic structural diagram of another test question generating system based on a knowledge graph and a prediction model according to an embodiment of the present invention. As shown in fig. 3, the system may include:
a memory 301 storing executable program code;
a processor 302 coupled with the memory 301;
the processor 302 invokes the executable program code stored in the memory 301 to perform some or all of the steps in the test question generation method based on the knowledge graph and the predictive model disclosed in the embodiment of the invention.
In a fourth embodiment, the present invention discloses a computer storage medium, where computer instructions are stored, and when the computer instructions are called, the computer instructions are used to execute part or all of the steps in the test question generating method based on the knowledge graph and the prediction model disclosed in the first embodiment of the present invention.
The system embodiments described above are merely illustrative, in which the modules illustrated as separate components may or may not be physically separate, and the components shown as modules may or may not be physical, i.e., may be located in one place, or may be distributed over a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above detailed description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course by means of hardware. Based on such understanding, the foregoing technical solutions may be embodied essentially or in part in the form of a software product that may be stored in a computer-readable storage medium including Read-Only Memory (ROM), random-access Memory (Random Access Memory, RAM), programmable Read-Only Memory (Programmable Read-Only Memory, PROM), erasable programmable Read-Only Memory (Erasable Programmable Read Only Memory, EPROM), one-time programmable Read-Only Memory (OTPROM), electrically erasable programmable Read-Only Memory (EEPROM), compact disc Read-Only Memory (Compact Disc Read-Only Memory, CD-ROM) or other optical disc Memory, magnetic disc Memory, tape Memory, or any other medium that can be used for computer-readable carrying or storing data.
Finally, it should be noted that: the embodiment of the invention discloses a test question generation method and a test question generation system based on a knowledge graph and a prediction model, which are disclosed by the embodiment of the invention only as a preferred embodiment of the invention, and are only used for illustrating the technical scheme of the invention, but not limiting the technical scheme; although the invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand that; the technical scheme recorded in the various embodiments can be modified or part of technical features in the technical scheme can be replaced equivalently; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.