Online examination automatic test paper composition method based on knowledge graph

Abstract

The invention belongs to the field of examination systems, and particularly relates to an online examination automatic test paper composition method based on a knowledge graph, and the method comprises the following steps: S1, data collection; s2, constructing an atlas; s3, collecting and analyzing examinee information; s4, analyzing examination requirements; s5, formulating a test paper generation strategy; and S6, test paper generation and verification. According to the online examination automatic test paper composition method based on the knowledge graph, knowledge materials are widely collected through multiple channels, data mining and semantic network technologies are used for carefully analyzing knowledge point relations, and the constructed knowledge graph is distinct in hierarchy and rich in semantics. And formulating a strategy by using an intelligent algorithm according to knowledge level evaluation and examination requirements of examinees during paper composition. For weak or important examination knowledge points of examinees, the number of questions, score weights and difficulty levels can be accurately controlled, blindness of traditional test paper composition is avoided, test paper is made to closely fit the reality of the examinees, and the knowledge mastering and application ability of the examinees is accurately evaluated.

Description

Technical Field

[0001] The present invention relates to the technical field of examination systems, and in particular to an automatic test paper generation method for online examinations based on a knowledge graph. Background Art

[0002] In the field of education, examinations are the most important part of the entire teaching process. Through examinations, we can have an accurate assessment of students' learning ability and the degree of their knowledge mastery, which helps parents and teachers to grasp students' learning status in a timely manner and provide guidance. In the current booming development of digital education, online examination systems are widely used. Their efficient organization and management, accurate data statistical analysis and strong resource integration capabilities provide a more fair, objective and efficient way for talent selection. In addition, online examination systems can also seamlessly connect with various teaching platforms and learning resources to form a complete educational ecosystem, realize the integration of teaching, learning and examination, and promote the innovative development of education and teaching models.

[0003] At present, the existing online test paper-forming devices lack the ability to deeply understand and accurately grasp the knowledge system when forming papers. They often simply extract questions based on question types and knowledge point distribution templates, without deeply exploring the internal logical relationship of knowledge points, and are unable to build a comprehensive and accurate knowledge graph. For example, in some subject exams, the relationship between knowledge points cannot be effectively identified, resulting in a lack of systematic and targeted test paper formation. For candidates with different professional directions, it is difficult to customize personalized test papers based on their knowledge weaknesses and key examination contents. The test papers have low differentiation and cannot accurately assess the candidate's ability level. This makes it difficult for the exam to accurately measure the candidate's true knowledge mastery and cannot provide a reliable basis for teaching and talent selection. In view of this, we propose an automatic test paper-forming method for online exams based on knowledge graphs. Summary of the invention

[0004] The main purpose of the present invention is to provide an automatic test paper generation method for online examinations based on knowledge graphs, which can solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above-mentioned purpose, the present invention proposes an automatic test paper generation method for online examinations based on knowledge graph, which comprises the following steps:

[0006] S1, data collection;

[0007] S2, map construction;

[0008] S3. Collection and analysis of candidate information;

[0009] S4. Exam requirements analysis;

[0010] S5. Formulate test paper composition strategy;

[0011] S6. Test paper generation and verification.

[0012] Preferably, in step S1, data can be widely collected from multiple sources such as authoritative textbooks, teaching outlines, academic research reports, professional course video explanations, etc., and the collected data can be preliminarily screened and sorted. For example, the computer science professional examination needs to cover programming languages, data structures, algorithm design, software engineering and other materials, and then remove redundant and erroneous information, and extract key knowledge point information, including names, definitions, subject branches, related concepts, etc.

[0013] Preferably, in step S2, the graph construction adopts semantic network technology to semantically annotate knowledge points and perform relationship modeling to construct a first knowledge graph with a hierarchical structure and semantic richness. Taking the chemical knowledge graph as an example, with the periodic table knowledge point as the core, the chemical reactions, chemical bonds, material structures and other knowledge points are associated to clearly display the chemical knowledge system architecture.

[0014] Preferably, in step S3, the candidate learning management system and the online learning platform are connected to obtain the candidate's comprehensive learning data and construct a second knowledge graph, including the completion of daily homework, periodic test scores, online learning time, learning resource click frequency and stay time, questions and interactive records during the learning process, etc. For example, for language learning tests, the candidate's vocabulary recitation software learning progress, oral practice time, and grammar learning course homework accuracy data are obtained. The second knowledge graph clearly shows the student's knowledge structure and the history of completed learning, homework, and exams.

[0015] Preferably, in step S3, a classification algorithm in machine learning is used to construct a knowledge level assessment model based on the second knowledge graph and the candidate's learning data, and the candidate's mastery of each knowledge point is divided into multiple levels such as proficient, good, medium, weak, and not mastered. Taking the engineering mechanics test as an example, the candidate's performance in the application of mechanics principles and force analysis questions and the learning process data are combined to determine their mastery level of knowledge points such as "static equilibrium" and "dynamic equations", so as to provide a personalized basis for test paper composition.

[0016] Preferably, in step S4, the detailed examination requirements input by the examination organizer through a special examination setting interface include the total examination duration, the expected total score of the examination paper, the number range and score setting of various question types, the coverage requirements of specific knowledge points and the emphasis level marking, etc. For example, the medical qualification examination has an examination duration of 180 minutes and a total score of 300 points. It stipulates 100 single-choice questions (1 point each), 50 multiple-choice questions (2 points each), 10 short-answer questions (10 points each), and 5 calculation questions (20 points each), focusing on covering core subject knowledge such as internal medicine, surgery, and physiology, and setting higher weights for key disease diagnosis and treatment knowledge points.

[0017] Preferably, in step S5, an intelligent optimization algorithm is used to formulate a personalized test paper composition strategy based on the candidate's knowledge level assessment results and the test requirement analysis data, and an intelligent optimization algorithm (such as a genetic algorithm, a particle swarm optimization algorithm, etc.) and the node correlation of the first and second graphs (such as the semantic-based node cosine similarity distance) are used to formulate a personalized test paper composition strategy based on the candidate's knowledge level assessment results and the test requirement analysis data.

[0018] Preferably, in step S6, according to the test paper composition strategy, questions are extracted from the carefully labeled and classified test question bank, and the difficulty coefficient and discrimination coefficient of the test paper need to be considered to ensure that the generated test paper can adapt to the overall level of the examinees and effectively distinguish examinees of different ability levels. The test questions in the test question bank are detailed and classified according to multi-dimensional attributes such as knowledge points, question types, and difficulty levels. For example, in the physics test question bank, mechanics questions are subdivided according to knowledge points such as Newton's laws of motion and the law of conservation of energy, and each knowledge point is further divided into easy, medium, and difficult levels, and the question type is marked.

[0019] Preferably, in step S6, the generated test paper is comprehensively checked to check whether the total score of the test paper, the number of question types and score distribution, and the knowledge point coverage meet the set requirements, etc. At the same time, the difficulty coefficient and discrimination coefficient of the test paper need to be considered, wherein the calculation formula of the test paper difficulty coefficient P is:

[0020]

[0021] The difficulty coefficient P of the test paper is obtained by taking the weighted sum of the difficulty coefficient Pi of each test question (which can be set in advance according to factors such as the depth of the knowledge point of the test question, the complexity of the problem-solving steps, etc.) and its full score Wi, and dividing it by the total number of test questions W (i.e. N). For example, the mathematics test verifies whether the total score of the test paper is 150 points, whether there are 30 multiple-choice questions and the score of each question is correct, whether it covers the predetermined knowledge point range and whether there is sufficient depth in testing the key functions and geometric figures knowledge points.

[0022] Preferably, in step S6, if a problem is found during verification, an adjustment mechanism is automatically started, and the test paper is optimized according to the type and severity of the problem by replacing questions, adjusting question scores, increasing or decreasing the number of questions, etc., until the requirements are met. For example, if it is found that the score of a question on a certain knowledge point is too high and the total score is out of range, a question with a similar difficulty level and a lower score to the knowledge point is selected to replace it or reduce the score of the question, so as to ensure the rationality and scientificity of the test paper.

[0023] The present invention provides an automatic test paper generation method for online examinations based on knowledge graphs. It has the following beneficial effects:

[0024] (1) This method of automatic test paper generation based on knowledge graph collects knowledge materials through multiple channels, uses data mining and semantic network technology to analyze the connection between knowledge points in detail, and constructs a first knowledge graph with clear hierarchy and rich semantics. When generating the test paper, the intelligent algorithm and the node correlation of the first and second graphs are used to formulate strategies based on the knowledge level assessment of the candidates and the test requirements. For the weak or key knowledge points of the candidates, the number of questions, score weights and difficulty levels can be accurately controlled, avoiding the blindness of traditional test paper generation, making the test paper closely fit the actual situation of the candidates, and accurately assessing the candidates' knowledge mastery and application ability.

[0025] (2) In terms of candidate information collection, this method for automatic online exam composition based on knowledge graph deeply connects with multiple platforms, automatically obtains and effectively manages and analyzes learning data, and constructs a second knowledge graph to represent students' knowledge structure and historical situations of completing learning, homework, and exams, providing timely and accurate data support for the generation of composition strategies. The paper generation and verification process relies on carefully planned processes and efficient algorithms, such as using inverted index technology to quickly retrieve questions and quickly verify and adjust according to preset rules and algorithms. The entire process does not require much manual operation, is fast and smooth, and effectively saves time. It perfectly meets the needs of fast composition for large-scale online exams and significantly improves the efficiency of composition.

[0026] (3) This method of automatic test paper generation based on knowledge graph comprehensively collects the learning data of candidates, builds a knowledge level assessment model with the help of the second knowledge graph and machine learning algorithm, accurately divides the knowledge points mastered by candidates, provides a personalized basis for test paper generation, and ensures that the test paper reflects the true level of candidates. The test paper generation and verification process strictly follows multi-dimensional standards such as knowledge point coverage, question type ratio, difficulty coefficient, and discrimination coefficient, reasonably arranges questions of different difficulty and knowledge points, and accurately verifies various indicators. The test paper generated in this way can effectively distinguish candidates of different levels, provide a strong evaluation basis for education and teaching and talent selection, and effectively promote the improvement of teaching quality and accurate talent screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0028] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0029] Figure 2 This is a schematic diagram of step S1 of the present invention;

[0030] Figure 3 This is a schematic diagram of step S2 of the present invention;

[0031] Figure 4 This is a schematic diagram of step S3 of the present invention;

[0032] Figure 5 This is a schematic diagram of step S4 of the present invention;

[0033] Figure 6 This is a schematic diagram of step S5 of the present invention;

[0034] Figure 7 This is a schematic diagram of step S6 of the present invention.

[0035] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] See also Figure 1-Figure 7 The present invention proposes an automatic test paper generation method for online examinations based on knowledge graph, comprising the following steps:

[0038] S1, data collection;

[0039] Among them, data can be widely collected from multiple sources such as authoritative textbooks, teaching outlines, academic research reports, and professional course video explanations. Knowledge materials related to the exam are then preliminarily screened and sorted. For example, computer science professional exams need to cover programming languages, data structures, algorithm design, software engineering, and other materials. Then, redundant and erroneous information is removed, and key knowledge points are extracted, including names, definitions, subject branches, and related concepts. Web crawler technology and data import tools are used to collect massive knowledge material data from multiple source platforms such as major educational resource websites, academic databases, and internal materials of educational institutions. For example, in the construction of financial exam knowledge graphs, data such as financial market analysis reports, financial regulations, and various financial investment course materials are collected. At the same time, lexical analysis and syntactic analysis in natural language processing technology are used to clean and preprocess the collected data to extract key knowledge points and semantic relationship information. For example, when processing financial regulations, core financial concepts, scope of application, constraints, and other information are extracted, and their correlation with other financial knowledge points is determined, such as the constraint and guidance relationship between a certain financial regulatory law and the risk management knowledge point of a financial institution.

[0040] S2, map construction;

[0041] Among them, in step S2, the graph construction uses semantic network technology to semantically annotate knowledge points and model relationships, and constructs a first knowledge graph with a hierarchical structure and semantic richness. Taking the chemical knowledge graph as an example, with the periodic table knowledge point as the core, the chemical reaction, chemical bond, material structure and other knowledge points are associated to show the chemical knowledge system architecture in a hierarchical manner. When the knowledge point information processed in step S1 is input into the graph database management system, the knowledge graph is constructed using the node and edge construction function, and the structure is continuously optimized, constraints and indexes are added, and query and reasoning efficiency are improved. For example, when constructing an economics knowledge graph, macroeconomics and microeconomics are used as the top-level nodes, and the knowledge point nodes such as economic theory, phenomenon, and policy are subdivided, and edges such as causal relationships and influence relationships are constructed to form a complete architecture.

[0042] S3. Collection and analysis of candidate information;

[0043] Among them, in step S3, the candidate learning management system and the online learning platform are connected to obtain the candidate's comprehensive learning data, and a second knowledge graph is constructed, including the completion of daily homework, periodic test scores, online learning time, learning resource click frequency and stay time, questions and interactive records during the learning process, etc. The second knowledge graph clearly shows the student's knowledge structure and the completion of learning, homework, exams and other historical situations. For example, for language learning exams, obtain data such as the candidate's vocabulary recitation software learning progress, oral practice time and grammar learning course homework accuracy, develop a dedicated data interface with the candidate's learning management system and online learning platform, use encrypted transmission protocols and data verification mechanisms to ensure data transmission stability and security, prevent data tampering or leakage, and regularly (such as 2 a.m. every day) obtain candidate learning data from the data interface, store it in a dedicated candidate data warehouse for centralized management and backup, and store it in the corresponding database table;

[0044] In step S3, a classification algorithm in machine learning is used to construct a knowledge level assessment model based on the second knowledge graph and the candidate's learning data, and the candidate's mastery of each knowledge point is divided into multiple levels such as proficient, good, medium, weak, and not mastered. Taking the engineering mechanics test as an example, the candidate's mastery level of knowledge points such as "static equilibrium" and "dynamic equations" is determined in combination with the candidate's performance in the application of mechanical principles and force analysis questions and the learning process data, providing a personalized basis for the test paper, cleaning and feature extraction of the data in the candidate's data warehouse, and converting the original data into feature vectors that can be used by the machine learning model. For example, the homework score is converted into a score rate feature, the study time is converted into a learning intensity feature, etc., and the candidate data is analyzed and evaluated using the trained knowledge level assessment model, and the assessment results are stored in the candidate's test archive for the test paper strategy generation module to call. Before each test, the system automatically obtains the candidate's latest knowledge level assessment results as the basis for test paper composition.

[0045] S4. Exam requirements analysis;

[0046] Among them, in step S4, the detailed examination requirements entered by the examination organizer through a special examination setting interface include the total examination duration, the expected total score of the examination paper, the number range and score setting of various question types, the coverage requirements of specific knowledge points and the emphasis level marking, etc. For example, the medical qualification examination has an examination duration of 180 minutes and a total score of 300 points. It stipulates 100 single-choice questions (1 point each), 50 multiple-choice questions (2 points each), 10 short-answer questions (10 points each), and 5 calculation questions (20 points each). It focuses on covering core subject knowledge such as internal medicine, surgery, and physiology, and sets a higher weight for key disease diagnosis and treatment knowledge points. The examination organizer enters detailed examination requirement information through the secure login examination setting management interface, and the system verifies and formats the input information in real time, such as checking whether the sum of the question type ratios is 100%. If it does not match, it prompts to re-enter to ensure that the input information is accurate.

[0047] S5. Formulate test paper composition strategy;

[0048] Among them, in step S5, according to the evaluation results of the candidate's knowledge level and the test demand analysis data, the intelligent optimization algorithm is used to formulate a personalized test paper composition strategy. According to the evaluation results of the candidate's knowledge level and the test demand analysis data, the intelligent optimization algorithm (such as genetic algorithm, particle swarm optimization algorithm, etc.) and the node correlation of the first and second graphs (such as the node cosine similarity distance based on semantics) are used to formulate a personalized test paper composition strategy. The system integrates the test requirement information and the candidate's knowledge level evaluation results and inputs them into the test paper composition strategy generation algorithm module. The test paper strategy generation algorithm module uses a genetic algorithm to perform multi-generation evolution optimization. In the first generation population, multiple initial test paper schemes are randomly generated, and the next generation scheme is generated through genetic operations such as crossover and mutation. The fitness value is calculated according to the optimization objectives such as the completeness of the test paper knowledge point coverage, the rationality of the question type distribution, and the adaptability to the candidate's knowledge level. The scheme with high fitness is selected to enter the next generation of evolution, and the optimal test paper composition strategy is obtained after multiple generations of evolution.

[0049] S6, test paper generation and verification;

[0050] Among them, in step S6, according to the test paper composition strategy, questions are extracted from the test question bank that has been carefully labeled and classified and stored. At the same time, the difficulty coefficient and discrimination coefficient of the test paper need to be considered to ensure that the generated test paper can not only adapt to the overall level of the candidates, but also effectively distinguish candidates with different ability levels. The questions in the test question bank are detailedly labeled and classified according to multi-dimensional attributes such as knowledge points, question types, and difficulty levels. For example, in the physics test bank, mechanics questions are subdivided according to knowledge points such as Newton's laws of motion and the law of conservation of energy. Each knowledge point is further divided into easy, medium and difficult levels, and the question type is marked. The test paper generation submodule uses the efficient index structure and query algorithm (such as inverted index technology) of the test paper bank management system based on the test paper composition strategy to quickly retrieve questions according to attributes such as knowledge points, question types, and difficulty levels; the extracted questions are combined into test papers in a predetermined order, and a unique identification code and encryption key are generated for the test paper to ensure its security and uniqueness; the generated test paper is fully verified to check whether the total score of the test paper, the number of question types and score distribution, and the knowledge point coverage meet the set requirements, etc. At the same time, the difficulty coefficient P and discrimination coefficient of the test paper need to be considered. The calculation formula of the test paper difficulty coefficient P is:

[0051]

[0052] The difficulty coefficient P of the test paper is obtained by weighted summing the difficulty coefficient Pi of each test question (which can be set in advance according to factors such as the depth of knowledge points of the test questions and the complexity of the problem-solving steps) and its full score Wi, and dividing it by the total number of test questions W (i.e. N). For example, the math test verifies whether the total score of the test paper is 150 points, whether there are 30 multiple-choice questions and the score of each question is correct, whether it covers the predetermined knowledge point range and has sufficient examination depth for key functions and geometric figures. The test paper verification submodule starts the verification program immediately after the test paper is generated, and checks each index of the test paper one by one according to the verification rules stored in the rule library; if the verification finds a problem, the adjustment submodule automatically adjusts according to the strategies in the adjustment strategy library (such as question replacement strategy, score adjustment strategy, etc.). After adjustment, it is verified again until the test paper meets the requirements and is stored in the test paper library to wait for the test to be issued.

[0053] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A method for automatically generating online examination papers based on knowledge graph, characterized in that: The following steps are involved: S1, data collection; S2, map construction; S3. Collection and analysis of candidate information; S4. Exam requirements analysis; S5. Formulate test paper composition strategy; S6. Test paper generation and verification.

2. According to claim 1, a method for automatically generating online examination papers based on knowledge graph, characterized in that: In step S1, data can be widely collected from multiple sources such as authoritative textbooks, teaching outlines, academic research reports, professional course video explanations, etc., and the collected data can be preliminarily screened and sorted.

3. According to claim 1, a method for automatically generating online examination papers based on knowledge graph, characterized in that: In step S2, the graph construction uses semantic network technology to semantically annotate knowledge points and perform relationship modeling to construct a first knowledge graph with a hierarchical structure and semantic richness.

4. According to claim 1, a method for automatically generating online examination papers based on knowledge graph, characterized in that: In step S3, the candidate learning management system and the online learning platform are connected to obtain the candidate's comprehensive learning data, and a second knowledge graph is constructed to represent the student's knowledge structure and the history of completed learning, homework, examinations, etc.

5. According to claim 1, a method for automatically generating online examination papers based on knowledge graph, characterized in that: In step S3, a classification algorithm in machine learning is used to build a knowledge level assessment model based on the second knowledge graph and the candidate's learning data.

6. The method for automatically generating online examination papers based on knowledge graph according to claim 1, characterized in that: In step S4, the test organizer enters detailed test requirements through a special test setting interface, including the total test duration, expected total score of the test paper, the number range and score settings of various question types, specific knowledge point coverage requirements and emphasis marks, etc.

7. The method for automatically generating online examination papers based on knowledge graph according to claim 1, characterized in that: In step S5, based on the candidate's knowledge level assessment results and the test requirement analysis data, a personalized test paper composition strategy is formulated using an intelligent optimization algorithm and the node correlation between the first and second graphs.

8. The method for automatically composing online examination papers based on knowledge graph according to claim 1, characterized in that: In step S6, questions are extracted from a carefully labeled and classified question bank according to the test paper composition strategy.

9. The method for automatically generating online examination papers based on knowledge graph according to claim 1, characterized in that: In step S6, the generated test paper is comprehensively checked to check whether the total score, the number of question types and score distribution, and the knowledge point coverage of the test paper meet the set requirements. At the same time, the difficulty coefficient P of the test paper needs to be considered. The calculation formula of the test paper difficulty coefficient is: The difficulty coefficient P of the test paper is obtained by taking the weighted sum of the difficulty coefficient Pi of each test question (which can be set in advance based on factors such as the depth of the knowledge points of the test question, the complexity of the problem-solving steps, etc.) and its full score Wi, and dividing it by the total number of test questions W (i.e. N).

10. The method for automatically generating online examination papers based on knowledge graph according to claim 1, characterized in that: In step S6, if problems are found during verification, the adjustment mechanism is automatically activated to optimize the test paper by replacing questions, adjusting question scores, increasing or decreasing the number of questions, etc., according to the type and severity of the problem, until the requirements are met.

Images