Test question making method and system based on artificial intelligence

Abstract

The invention relates to the technical field of intelligent education, and discloses a test question making method and system based on artificial intelligence. The method comprises the following steps: firstly, constructing a digital resource library comprising a knowledge point library, a test question type feature library and a material library, wherein the knowledge point library comprises attribute information such as mastering requirements, difficulty degree, cognitive level and level depth of knowledge points; evaluating the question editing materials through three dimensions of investigation content adaptation degree, investigation question type adaptation degree and investigation target adaptation degree, and determining candidate question editing materials based on an evaluation result; designing a question compiling rule according to the question type features, and generating test questions based on the candidate question compiling materials and the question compiling rule; and finally, the generated test questions are subjected to digital coding and warehousing after being manually audited. According to the method, the test question quality and the teaching suitability are remarkably improved, the generated test questions meet the requirements of specific learning groups, the test question knowledge points, question types and difficulty requirements can be accurately matched, and the method has innovativeness and practicability.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to test question preparation technology based on artificial intelligence. Background Art

[0002] In the teaching process, test questions are an important tool for assessing student learning outcomes and consolidating knowledge application. Traditional test question creation relies primarily on manual compilation by teachers, which presents the following problems: First, manual compilation is time-consuming and labor-intensive, requiring teachers to devote considerable time and energy to collecting materials, designing question types, and organizing items, making it difficult to meet the preparation needs of daily teaching; second, the quality of manually compiled test questions varies, with a single type of question and a lack of specificity and systematicity, making it difficult to comprehensively and objectively assess students' mastery of knowledge points.

[0003] To improve the efficiency and quality of test question writing, some studies have proposed using computer technology to automatically generate test questions. For example, a common approach is to generate test questions based on templates and rules. Question templates and generation rules are pre-designed, and then relevant knowledge points and materials are extracted from a question bank to form test questions. However, this approach typically requires significant manual effort to design templates and rules, and the generated test questions lack flexibility, making them difficult to adapt to changing teaching needs.

[0004] Another approach uses natural language processing technology to automatically extract knowledge points from text materials and organize them into questions. However, this approach requires a high degree of structuring of the material and generates relatively limited question types, typically limited to multiple-choice and fill-in-the-blank questions, making it difficult to generate logically and comprehensively structured questions.

[0005] Furthermore, existing test question generation methods often suffer from a disconnect between the generated questions and teaching requirements. These methods lack specificity in terms of difficulty, knowledge coverage, and question type design, and fail to fully consider factors such as students' cognitive level and course progress. Consequently, the generated questions fail to fully meet the requirements of "teaching students in accordance with their aptitude."

[0006] In summary, how to achieve the rapid, accurate and large-scale generation of high-quality test questions, and thus effectively improve the efficiency of teachers' preparation and students' learning, is a key issue that needs to be urgently solved in the field of automatic test question generation. Summary of the Invention

[0007] The purpose of this application is to provide an artificial intelligence-based test question creation method and system to solve the problems raised in the above background technology.

[0008] This application discloses a test question preparation method based on artificial intelligence, comprising the following steps:

[0009] Build a digital resource library including a knowledge point library, a test question type feature library, and a material library. The knowledge point library contains attribute information such as the mastery requirements, difficulty level, cognitive level, and hierarchical depth of knowledge points. The test question type feature library contains question type feature information. The material library contains materials corresponding to knowledge points.

[0010] Select a target knowledge point from the knowledge point library as the knowledge point for question compilation and determine the question compilation type. Based on the material library, obtain the question compilation materials corresponding to the knowledge point for question compilation.

[0011] Evaluate the adaptability of the question compilation materials, including: calculating the adaptability of the inspected content, which represents the matching degree between the content of the question compilation materials and the knowledge point for question compilation; calculating the adaptability of the inspected question type, which represents the matching degree between the format of the question compilation materials and the question compilation type; calculating the adaptability of the inspected target, which represents the matching degree between the difficulty of the question compilation materials and the difficulty of the test questions; and calculating the comprehensive adaptability of the question compilation materials based on the adaptability of the inspected content, the adaptability of the inspected question type, and the adaptability of the inspected target by using a weighted calculation method.

[0012] Determine the question compilation materials with a comprehensive adaptability higher than the set threshold as candidate question compilation materials.

[0013] Based on the question type feature information in the test question type feature library, obtain question type features including the stem and answer generation requirements, and design question compilation rules according to the question type features.

[0014] Generate test questions based on the candidate question compilation materials and the question compilation rules.

[0015] Digitally encode the generated test questions after manual review and store them in the test question type feature library to complete the storage of test questions.

[0016] In a preferred example, the mastery requirements are divided into three levels: recognition and understanding, understanding and explanation, and mastery and application, with values of f′, f″, and f′′′ respectively, and 0 < f′ < f″ < f′′′ < 1; the difficulty level is divided into three levels: easy, medium, and difficult, with values of g′, g″, and g′′′ respectively, and 0 < g′ < g″ < g′′′ < 1; the cognitive level is determined by the average score of the teaching group answering test questions containing this knowledge point, and is divided into three levels: low, medium, and high, with values of h′, h″, and h′′′ respectively, and 0 < h′ < h″ < h′′′ < 1; the hierarchical depth refers to the depth of the three levels of the basic layer, associated layer, and application layer divided according to the complexity and logical relationship of the knowledge point content in the knowledge system, with values of k′, k″, and k′′′ respectively, and 0 < k′ < k″ < k′′′ < 1.

[0017] In a preferred embodiment, the adaptability of the inspection content is calculated by the following formula:

[0018]

[0019] Among them, Match(A,B) represents the matching degree between A and B; Min(a,b) is the total number of knowledge points contained in the question material, Max(a,b) is the total number of knowledge points contained in the test knowledge point domain, and min(a,b) is the average total number of knowledge points contained in this question type; X i 、X j are the assessment weights of the i-th and j-th knowledge points in the test material; Y i 、Y j are the requirements for mastering the i-th and j-th knowledge points in the question-making materials; Z i 、Z j are the cognitive levels of the teaching group for the i-th and j-th knowledge points in the question-making materials; α1, α2, α3, δ1, δ2, ε1, and ε2 are the weight coefficients respectively.

[0020] In a preferred example, the assessment weights include: assessment knowledge point weight value μ1=0.8, general knowledge point weight value μ2=0.2, where the assessment knowledge points are question-writing knowledge points, and general knowledge points are knowledge points other than question-writing knowledge points.

[0021] In a preferred embodiment, the suitability of the test question type is calculated by the following formula:

[0022]

[0023] Among them, Match(A,B) represents the matching degree between A and B; S is the type of the editing material, S′ is the type of the editing question type; P i 、P j is the type of knowledge point i and j in the question-making material; Q i , Q j are the depths of the i-th and j-th knowledge points in the question-making materials; β1, β2, β3, θ1, θ2, are weight coefficients respectively.

[0024] In a preferred embodiment, the target suitability is calculated by the following formula:

[0025]

[0026] Among them, Match(A,B) represents the matching degree between A and B; R is the average difficulty of each knowledge point in the question material, R′ is the average difficulty of the question type; R i 、R j are the difficulty levels of the i-th and j-th knowledge points in the question-making materials; Zi 、Z j are the cognitive levels of the teaching group for the i-th and j-th knowledge points in the question-making materials; γ1, γ2, γ3, ρ1, ρ2, σ1, and σ2 are the weight coefficients respectively.

[0027] In a preferred example, the difficulty of the test question is determined by the ratio of the number of people who answer the test question correctly to the total number of people, and the discrimination of the test question is determined by the difference between the ratio of the number of people who answer the test question correctly in the high-scoring group and the ratio of the number of people who answer the test question correctly in the low-scoring group.

[0028] In a preferred embodiment, the calculation formula of the comprehensive fitness is:

[0029] F=xf1+yf2+zf3

[0030] Among them, F is the comprehensive adaptability result of the test material, f1, f2, and f3 are the adaptability of the test content, the adaptability of the test question type, and the adaptability of the test goal respectively, and x, y, and z are the weight values ​​of the three dimensions respectively, and 0 <x<1,0<y<1,0<z<1,x+y+z=1。

[0031] In a preferred example, for multiple-choice questions, the question-writing rules include: based on the example sentence materials of the question-writing knowledge points, the target knowledge points are hollowed out through knowledge point recognition technology to generate the question stem; interference items are generated in priority order through the changed word forms, synonyms and antonyms, and phrase collocations of the target knowledge points; the target knowledge points and interference items that are hollowed out are randomly sorted to generate option groups and answers.

[0032] In a preferred example, for vocabulary fill-in-the-blank questions, the question-writing rules include: based on example sentence materials of the question-writing knowledge points, extracting the target knowledge points from them through knowledge point recognition technology, adding the first letter prompt information of the knowledge point while leaving blanks to generate the question stem; and directly using the extracted target knowledge points as the answer to the question.

[0033] In a preferred example, for translation questions, the question-making rules include: using example sentences of the question-making knowledge points as the question stem; and using the translation sentences corresponding to the example sentences as the answers to the questions.

[0034] This application also discloses an artificial intelligence-based test question making system, comprising:

[0035] A digital resource library construction module is used to construct a digital resource library including a knowledge point library, a test question type feature library, and a material library. The knowledge point library contains attribute information on the mastery requirements, difficulty level, cognitive level, and layer depth of the knowledge points. The test question type feature library contains question type feature information. The material library contains materials corresponding to the knowledge points.

[0036] A question writing material acquisition module is used to select target knowledge points from the knowledge point library as question writing knowledge points, determine the question writing type; based on the material library, obtain question writing materials corresponding to the question writing knowledge points;

[0037] The adaptability evaluation module is used to evaluate the adaptability of the editing material, including: calculating the adaptability of the examination content, the examination content adaptability represents the degree of matching between the content of the editing material and the editing knowledge points; calculating the adaptability of the examination question type, the examination question type adaptability represents the degree of matching between the format of the editing material and the editing question type; calculating the adaptability of the examination target, the examination target adaptability represents the degree of matching between the difficulty of the editing material and the difficulty of the test questions; and using a weighted calculation method to calculate the comprehensive adaptability of the editing material based on the examination content adaptability, the examination question type adaptability and the examination target adaptability;

[0038] A candidate material determination module is used to determine the editing materials with comprehensive adaptability higher than a set threshold as candidate editing materials;

[0039] A question editing rule design module is used to obtain question type features including question stems and answer generation requirements based on question type feature information in the question type feature library, and design question editing rules according to the question type features;

[0040] A test question generating module, configured to generate test questions based on the candidate test question materials and the test question editing rules;

[0041] The test question storage module is used to digitally encode the generated test questions after manual review and store them in the test question type feature library to complete the test question storage.

[0042] The present application provides an artificial intelligence-based test question creation method, which constructs digital resources such as a knowledge point library, a test question type feature library, and a material library, and selects target knowledge points from the knowledge point library as question-writing knowledge points to obtain question-writing materials corresponding to the question-writing knowledge points.

[0043] The suitability of the test materials is assessed, including suitability for the test content, test question types, and test objectives. The test content suitability calculation incorporates factors such as the number of knowledge points contained in the test materials and the test knowledge point domain, the average number of knowledge points, the assessment weight of each knowledge point, mastery requirements, and the cognitive level of the teaching group. The test question type suitability calculation incorporates factors such as the type match between the test materials and the question type, the match between the type and depth of each knowledge point and the question type. The test objective suitability calculation incorporates factors such as the match between the average difficulty level of each knowledge point in the test materials and the question type, the difficulty level of each knowledge point, and the cognitive level of the teaching group. By weightedly calculating each suitability, the comprehensive suitability of the test materials is obtained, and candidate test materials whose suitability meets the threshold are screened out.

[0044] Furthermore, based on the question type feature information in the question type feature library, the design requirements for the question stem, answer, and options are obtained, and the question editing rules are designed accordingly. For example, for multiple-choice questions, key knowledge points are hollowed out from the example sentence materials through knowledge point recognition technology to form the question stem, and interference items are generated through the variations of knowledge points, synonyms and antonyms, collocations, etc., and randomly arranged with the correct answers to form options. For fill-in-the-blank questions, target knowledge points are extracted from the example sentence materials through knowledge point recognition technology, leaving them blank and retaining the first letters, and the extracted knowledge points are used as the answers. For translation questions, the example sentence materials and translations of the knowledge points are directly used as the question stem and answer, respectively.

[0045] Finally, test questions are automatically generated based on the candidate materials and the rules for question creation. After expert review and approval, standardized test questions are formed. Simultaneously, the generated test questions are digitally encoded and stored in a question type feature database, enabling standardized test question management.

[0046] This application constructs a detailed attribute portrait from multiple dimensions such as the requirements for mastering knowledge points, difficulty, cognitive level, and depth of levels, and evaluates the adaptability of the editing materials from three dimensions: assessment content, question type characteristics, and target difficulty, thereby improving the matching degree between the test content and teaching requirements. At the same time, corresponding editing rules are designed based on the characteristics of different question types, realizing the automatic generation and standardization of test questions. Compared with existing technologies, this application can generate high-quality test questions quickly, accurately, and on a large scale, reducing the burden of teacher preparation and improving the quality of test questions and teaching effectiveness.

[0047] The specification of this application records a large number of technical features, which are distributed in various technical solutions. If all possible combinations of technical features of this application (i.e., technical solutions) are to be listed, the specification will be too lengthy. In order to avoid this problem, the various technical features disclosed in the above-mentioned invention content of this application, the various technical features disclosed in the various embodiments and examples below, and the various technical features disclosed in the accompanying drawings can be freely combined with each other to form various new technical solutions (these technical solutions are all deemed to have been recorded in this specification), unless such a combination of technical features is technically infeasible. For example, in one example, feature A+B+C is disclosed, and in another example, feature A+B+D+E is disclosed. Features C and D are equivalent technical means that play the same role. Technically, only one of them can be used, and it is impossible to use them at the same time. Feature E can be technically combined with feature C. Then, the solution of A+B+C+D should not be considered as having been recorded because it is technically infeasible, while the solution of A+B+C+E should be considered as having been recorded. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1It is a schematic diagram of a test question making system in the test question making method and system based on artificial intelligence according to the present application.

[0049] Figure 2 It is a mind map for determining the materials for writing questions in the artificial intelligence-based test question making method and system of this application.

[0050] Figure 3 It is a flowchart of an artificial intelligence-based test question preparation method according to an embodiment of the present application.

[0051] Figure 4 It is a structural diagram of an artificial intelligence-based test question preparation system according to an embodiment of the present application. DETAILED DESCRIPTION

[0052] In the following description, many technical details are provided to help readers better understand this application. However, those skilled in the art will understand that even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in this application can be implemented.

[0053] Description of some concepts:

[0054] Knowledge Points: In this application, knowledge points refer to the fundamental concepts, principles, theorems, and formulas within a discipline, and are the basic units that constitute the disciplinary knowledge system. Each knowledge point can be composed of elements such as name, definition, and attributes. For example, in mathematics, "function" is a knowledge point, including attributes such as domain, range, and function expression.

[0055] Knowledge Point Database: The knowledge point database in this application refers to a database that systematically stores subject knowledge points in digital form. The knowledge point database not only stores basic information about the knowledge points, but also includes attributes such as the relationships between the knowledge points, the requirements for mastering the knowledge points, their difficulty level, cognitive level, and hierarchical depth. These attributes together form a multi-dimensional portrait of the knowledge points.

[0056] Question Type Feature Library: The question type feature library in this application refers to a database that stores question type templates and features. This includes the structured definition of the question type, question stem and footer templates, and answer and option generation rules. These features collectively characterize the basic attributes and generation requirements of each question type. For example, the question type features of a multiple-choice question may include the question stem, number of options, and distractor generation rules.

[0057] Material Library: The material library in this application refers to a database that stores various materials used to generate test questions. These materials can include text, images, audio, and video, and are used to support the context setting, question formulation, and option generation of test questions. The materials in the material library are associated with knowledge points, allowing for quick retrieval of relevant materials through the knowledge point index.

[0058] Question-writing materials: Question-writing materials in this application refer to materials retrieved from the material library that are relevant to the knowledge points of the test questions to be generated. Question-writing materials are the direct raw materials for generating test questions, and their content quality and relevance directly affect the quality of the generated test questions. This application selects high-quality question-writing materials through a multi-dimensional suitability assessment of the materials.

[0059] Question type features: In this application, question type features refer to feature vectors that characterize the properties of the test question type. Question type features include the structured definition, components, and generation rules of the question type, and are used to guide the test question generation process. This application extracts question type features from a question type feature library and uses them to design corresponding question editing rules to generate test questions of a specific question type.

[0060] The following is a summary of some of the innovative features of this application:

[0061] In general, this application forms a refined representation of knowledge points by constructing a multi-dimensional attribute portrait of knowledge points in the knowledge point library, including mastery requirements (such as understanding f′, understanding explanation f″, mastering application f′'′), difficulty level (such as easy g′, medium g″, difficult g′'′), cognitive level (such as low h′, medium h″, high h′'′) and hierarchical depth (such as basic layer k′, association layer k″, application layer k′'′), etc. On this basis, for the knowledge points of the test questions to be generated, the corresponding question-making materials are obtained from the material library.

[0062] Furthermore, a multi-dimensional adaptation evaluation method is creatively proposed, which calculates the degree of match between the editing material and the test questions from three perspectives: examination content, question type characteristics, and target difficulty. The calculation of the examination content adaptation f1 not only considers the matching of the editing material and the test knowledge point domain in terms of the total number of knowledge points Min(a,b) / Max(a,b), but also takes into account the average number of knowledge points included min(a,b) / Max(a,b), the assessment weight of each knowledge point X i / X j , master the requirements Y i / Y j and cognitive level Z i / Z j The comprehensive content adaptation degree f1 is obtained by weighted summing the influencing factors ∑.

[0063] Similarly, in the calculation of the test question type adaptability f2 and the test target adaptability f3, the test material and the test question type are respectively based on the type S / S′ and the knowledge point type P. i / P j , layer depth Q i / Q j , as well as the average difficulty R / R′, the knowledge point difficulty Ri / R j , cognitive level Z i / Z j The matching degree Match(.) was evaluated in various aspects, and finally, through weighted summation ∑, the comprehensive question type suitability f2 and target suitability f3 were obtained. On this basis, the suitability of the three dimensions was linearly weighted using the form F = xf1 + yf2 + zf3 to obtain the comprehensive suitability F of the editing material.

[0064] Through the above-mentioned fitness evaluation, candidate editing materials with a comprehensive fitness F reaching the set threshold are screened out. For the candidate editing materials, this application further obtains question type features such as stem, answer and option design requirements that match the question type to be generated from the question type feature library, and designs a set of editing rules based on individual needs. For example, for multiple-choice questions, the knowledge points in the editing materials are hollowed out to form the stem through knowledge point recognition technology, and interference items are generated using the changing forms, synonyms and antonyms, and collocation words of the knowledge points, and randomly arranged with the correct answers into options and answers; for fill-in-the-blank questions, the editing materials are refined and hollowed out through knowledge point recognition technology, retaining the first letters, and the extracted knowledge points are used as answers; for translation questions, the examples and translations of the editing materials are directly used as the stem and answer respectively.

[0065] Finally, this application automatically generates the test questions to be generated based on the screened candidate question materials and designed question editing rules. After being reviewed and approved by experts, they are digitally encoded to form standardized test questions and stored in the test question type feature library for subsequent teaching use.

[0066] In summary, this application constructs a multi-dimensional attribute portrait of knowledge points, proposes a multi-angle and fine-grained method for evaluating the adaptability of editing materials, and specifically designs editing rules to form a complete set of automatic test question generation solutions. It organically combines and coordinates the knowledge point attribute portrait, the evaluation of the adaptability of editing materials, and the editing rules of individualized approach, overcoming the problems of the existing technology in that the generation of test questions is out of touch with the teaching requirements and the single question type, and realizes the rapid, accurate, and large-scale generation of high-quality test questions, effectively improving the efficiency of teachers preparing questions and students learning. This highlights the creative conception and significant technological progress of this application, which is not easily achieved by existing technologies.

[0067] The first embodiment of the present application relates to a test question making method based on artificial intelligence, the process of which is as follows: Figure 3 As shown, the method includes the following steps:

[0068] Step 100: Construct a digital resource library including a knowledge point library, a test question type feature library and a material library. The knowledge point library contains the mastery requirements, difficulty, cognitive level, and hierarchical depth attribute information of the knowledge points. The test question type feature library contains question type feature information. The material library contains materials corresponding to the knowledge points.

[0069] Step 200: Select a target knowledge point from the knowledge point library as a question-writing knowledge point and determine a question-writing type; based on the material library, obtain question-writing materials corresponding to the question-writing knowledge point.

[0070] Step 300: Evaluate the adaptability of the editing material, including: calculating the adaptability of the examination content, the examination content adaptability represents the degree of matching between the content of the editing material and the editing knowledge points; calculating the adaptability of the examination question type, the examination question type adaptability represents the degree of matching between the format of the editing material and the examination question type; calculating the adaptability of the examination target, the examination target adaptability represents the degree of matching between the difficulty of the editing material and the difficulty of the test questions; and using a weighted calculation method to calculate the comprehensive adaptability of the editing material based on the examination content adaptability, the examination question type adaptability and the examination target adaptability.

[0071] Step 400: Determine the editing materials with comprehensive adaptability higher than a set threshold as candidate editing materials.

[0072] Step 500: Based on the question type feature information in the test question type feature library, obtain question type features including question stem and answer generation requirements, and design question editing rules according to the question type features.

[0073] Step 600: Generate test questions based on the candidate test materials and the test rules;

[0074] Step 700: The generated test questions are manually reviewed, digitally encoded, and stored in the test question type feature database, completing the storage of the test questions.

[0075] Optionally, in step 100, the mastery requirements are divided into three levels: acquaintance, understanding, and application, with the values being f′, f″, and f′′′ respectively, and 0 < f′ < f″ < f′′′ < 1; the difficulty levels are divided into three levels: easy, medium, and difficult, with the values being g′, g″, and g′′′ respectively, and 0 < g′ < g″ < g′′′ < 1; the cognitive level is determined by the average score of the teaching group in answering the questions containing this knowledge point, and is divided into three levels: low, medium, and high, with the values being h′, h″, and h′′′ respectively, and 0 < h′ < h″ < h′′′ < 1; the level of depth refers to the depth of the three levels of basic layer, associated layer, and application layer divided according to the complexity and logical relationship of the knowledge point content in the knowledge system, with the values being k′, k″, and k′′′ respectively, and 0 < k′ < k″ < k′′′ < 1.

[0076] Specifically, the present application constructs a digital resource library including a knowledge point library, a test question type feature library, and a material library, providing comprehensive resource support for test question production. Based on the determined question-making knowledge points and question types, the optimal question-making materials are determined through multi-dimensional fitness evaluation, and then the corresponding question-making rules are designed according to the characteristics of different question types, and finally high-quality test questions that meet the needs of a specific learning group and accurately match the knowledge points, question types, and difficulty requirements are generated. This method significantly improves the quality of test questions and teaching adaptability, and realizes the intelligent and large-scale generation of test questions.

[0077] More specifically, in step 100, the knowledge point library makes a refined portrait of the knowledge points through multi-dimensional attribute information. Among them, the mastery requirements reflect the cognitive level that learners should reach for the knowledge points, from superficial acquaintance to deep application, reflecting the progression of learning; the difficulty level objectively reflects the complexity of the knowledge points themselves; the cognitive level is determined based on the actual performance of the teaching group, reflecting the actual mastery state of the learners; the level of depth divides the position of the knowledge points in the entire subject framework from the perspective of the knowledge system. The coordinated cooperation of these attribute information provides an accurate reference basis for the subsequent material fitness evaluation, ensuring a high degree of matching between test question generation and teaching requirements.

[0078] More specifically, in step 200, the target knowledge points are selected from the knowledge point library as the question-making knowledge points, and the corresponding question-making question types are determined at the same time. This step clarifies the core elements of test question generation. The question-making materials corresponding to the question-making knowledge points are obtained based on the material library, providing candidate materials for the subsequent fitness evaluation. The selection of the question-making knowledge points directly determines the key points of the test questions, and the determination of the question-making question types stipulates the manifestation form of the test questions. The combination of the two lays the basic framework for test question generation.

[0079] Optionally, in step 300, the content fitness of the examination is calculated by the following formula:

[0080]

[0081] Among them, Match(A,B) represents the matching degree between A and B; Min(a,b) is the total number of knowledge points contained in the question material, Max(a,b) is the total number of knowledge points contained in the test knowledge point domain, and min(a,b) is the average total number of knowledge points contained in this question type; X i 、X j are the assessment weights of the i-th and j-th knowledge points in the test material; Y i 、Y j are the requirements for mastering the i-th and j-th knowledge points in the question-making materials; Z i 、Z j are the cognitive levels of the teaching group for the i-th and j-th knowledge points in the question-making materials; α1, α2, α3, δ1, δ2, ε1, and ε2 are the weight coefficients respectively.

[0082] More specifically, the calculation of content fit comprehensively assesses the degree of alignment between the material and the knowledge points used. The first dimension evaluates the matching of the material's knowledge point coverage with the test requirements by measuring the proportional relationship between the number of knowledge points. The second dimension ensures that the importance of each knowledge point in the material aligns with the teaching requirements by measuring the matching index between assessment weights and mastery requirements. The third dimension ensures that the material's difficulty aligns with the learner's actual cognitive abilities by measuring the matching index between assessment weights and cognitive level. These three dimensions work together through a weighted summation method to comprehensively assess the material's content-level applicability.

[0083] Optionally, in step 300, the assessment weights include: assessment knowledge point weight value μ1=0.8, general knowledge point weight value μ2=0.2, where the assessment knowledge points are question-writing knowledge points, and general knowledge points are knowledge points other than question-writing knowledge points.

[0084] Optionally, in step 300, the suitability of the examination question type is calculated using the following formula:

[0085]

[0086] Among them, Match(A,B) represents the matching degree between A and B; S is the type of the editing material, S′ is the type of the editing question type; P i 、P j is the type of knowledge point i and j in the question-making material; Q i , Q j are the depths of the i-th and j-th knowledge points in the question-making materials; β1, β2, β3, θ1, θ2, are weight coefficients respectively.

[0087] More specifically, the suitability of the examination question type is assessed from the perspective of the matching of the material format with the requirements of the question type. The first dimension directly compares the type of the question-making material with the degree of match between the type of question and the type of question, ensuring that the material's presentation can support the generation requirements of the corresponding question type. The second dimension evaluates the matching between the type of knowledge point and the question type, ensuring that the types of knowledge points contained in the material are suitable for the examination method of the target question type. The third dimension analyzes the matching between the depth of the knowledge point level and the question type, ensuring that the knowledge level of the material matches the depth of the examination of the question type. The coordinated evaluation of these three dimensions effectively ensures that the format and structure of the question-making material are highly compatible with the target question type.

[0088] Optionally, in step 300, the inspection target fitness is calculated using the following formula:

[0089]

[0090] Among them, Match(A,B) represents the matching degree between A and B; R is the average difficulty of each knowledge point in the question material, R′ is the average difficulty of the question type; R i 、R j are the difficulty levels of the i-th and j-th knowledge points in the question-making materials; Z i 、Z j are the cognitive levels of the teaching group for the i-th and j-th knowledge points in the question-making materials; γ1, γ2, γ3, ρ1, ρ2, σ1, and σ2 are the weight coefficients respectively.

[0091] More specifically, the target fit assessment focuses on evaluating the degree to which the difficulty of the material matches the target difficulty of the test questions. The first dimension compares the overall difficulty of the material with the average difficulty of the question type to ensure that the difficulty of the generated test questions meets the expected requirements. The second dimension assesses the matching of the difficulty of each knowledge point with the difficulty of the question type to ensure that the difficulty of each knowledge point in the material is appropriate to the target question type. The third dimension analyzes the matching of the difficulty of the knowledge point with the cognitive level of the teaching group to ensure that the difficulty of the test questions is consistent with the actual ability level of the learners. This comprehensive assessment of these three dimensions effectively ensures the accuracy of the difficulty control of the generated test questions.

[0092] Optionally, in step 300, the difficulty of the test question is determined by the ratio of the number of people who answer the test question correctly to the total number of people, and the discrimination of the test question is determined by the difference between the ratio of the number of people who answer the test question correctly in the high-scoring group and the ratio of the number of people who answer the test question correctly in the low-scoring group.

[0093] Optionally, in step 300, the calculation formula of the comprehensive fitness is:

[0094] F=xf1+yf2+zf3

[0095] Among them, F is the comprehensive adaptability result of the test material, f1, f2, and f3 are the adaptability of the test content, the adaptability of the test question type, and the adaptability of the test goal respectively, and x, y, and z are the weight values ​​of the three dimensions respectively, and 0 <x<1,0<y<1,0<z<1,x+y+z=1。

[0096] More specifically, the comprehensive adaptability evaluation combines the adaptability results of the three dimensions using a linear weighting method to form a comprehensive assessment of the overall suitability of the test material. The weights x, y, and z can be adjusted based on different application scenarios and needs. For example, if content matching is important, the value of x can be increased, while if question format is important, the value of y can be increased. This flexible weighting mechanism allows the adaptability evaluation to adapt to different teaching needs and test question generation objectives, while avoiding the one-sidedness that can arise from single-dimensional evaluation through comprehensive assessment.

[0097] More specifically, in step 400, by setting a comprehensive fitness threshold, question-writing materials with evaluation results above the threshold are screened as candidate question-writing materials, achieving intelligent screening of materials. This threshold is set based on actual teaching needs and test quality requirements. A threshold that is too high may result in too few available materials, while a threshold that is too low may affect test quality. The determination of candidate question-writing materials provides high-quality raw materials for subsequent test question generation, ensuring that the generated test questions meet the preset teaching objectives and quality standards in terms of content, format, and difficulty.

[0098] Optionally, in step 500, for multiple-choice questions, the question-writing rules include: based on the example sentence materials of the question-writing knowledge points, hollowing out the target knowledge points through knowledge point recognition technology to generate the question stem; generating interference items in priority order through the changed word forms, synonyms and antonyms, and phrase collocations of the target knowledge points; and randomly sorting the target knowledge points and interference items that are hollowed out to generate option groups and answers.

[0099] More specifically, the rules for multiple-choice questions have been meticulously designed to address the characteristics of this question type. Knowledge point recognition technology accurately locates and refines target knowledge points within example sentences, creating a clearly defined question stem. Distractors are generated by using various variations of the target knowledge point, including inflectional variations, semantically similar or opposite words, and common phrase combinations. These are prioritized based on their similarity to the correct answer and their level of interference, ensuring that the distractors are both deceptive and not overly simplistic. The random ordering of the option groups avoids patterned answer placement, enhancing the scientific nature and fairness of the test.

[0100] Optionally, in step 500, for vocabulary fill-in-the-blank questions, the question-writing rules include: based on example sentence materials of question-writing knowledge points, extracting target knowledge points from them through knowledge point recognition technology, adding the first letter prompt information of the knowledge point while leaving blanks to generate the question stem; and directly using the extracted target knowledge point as the answer to the question.

[0101] More specifically, the rules for vocabulary fill-in-the-blank questions fully consider the specific nature of this question type, which tests vocabulary proficiency. Knowledge point recognition technology not only accurately extracts the target knowledge points, but also retains appropriate hints, such as initial letters, when removing blanks. This maintains a certain level of difficulty while providing learners with necessary clues. The blank spaces are carefully designed to ensure that the context provides sufficient semantic support for the answer. By directly using the extracted target knowledge points as the answers, the accuracy and uniqueness of the answers are guaranteed, meeting the standardized requirements for vocabulary fill-in-the-blank questions.

[0102] Optionally, in step 500, for translation questions, the question-editing rules include: using example sentence materials of the question-editing knowledge points as the question stem; and using the translation sentence corresponding to the example sentence as the answer to the question.

[0103] More specifically, the rules for the translation questions reflect the test's emphasis on language conversion skills. The example sentences used directly as the stem preserve the integrity of the original text and the naturalness of the language, providing learners with authentic language application scenarios. The corresponding translated sentences serve as the standard answers, ensuring accuracy and authenticity. This direct correspondence simplifies the question-writing process while ensuring test quality, enabling translation questions to effectively assess learners' understanding and application of the proposed knowledge points in cross-language expression.

[0104] More specifically, the test item attributes are determined using scientific statistical analysis methods. The difficulty of the test items is quantitatively assessed using the correct answer rate. A higher correct answer rate indicates a lower difficulty level. This objective difficulty assessment avoids subjective bias. The discriminatory power of the test items is assessed by comparing the performance of learners with different ability levels. The greater the difference in correct answer rates between the high-performing and low-performing groups, the better the discriminatory power of the test items, effectively distinguishing between learners of different abilities. The coordinated application of these two attribute indicators provides a scientific basis for the assessment of test item quality and the subsequent management of the test item bank, ensuring the quality and applicability of the test items in the test item bank.

[0105] In order to better understand the technical solution of the present application, a specific example is provided below for illustration. The details listed in the example are mainly for ease of understanding and are not intended to limit the scope of protection of the present application.

[0106] Therefore, this example proposes an AI-based test question creation method and system. Based on the identified knowledge points and question types, this method determines the question-making materials by considering dimensions such as "test content suitability, test type suitability, and test objective suitability." Then, based on the design rules for different question types, it generates test questions that meet the needs of specific learning groups and precisely match the knowledge points, question types, and difficulty levels. This method significantly improves test question quality and teaching adaptability, and the generated questions are both innovative and practical.

[0107] This example proposes a test question production method and system based on artificial intelligence. The method first constructs a digital resource library including a knowledge point library, a test question type feature library, and a material library to provide resources for test question production; then designs an AI test question production engine, which first obtains the corresponding test material based on the test knowledge points, and then comprehensively evaluates the adaptability of the test material from three dimensions: test content adaptability, test question type adaptability, and test target adaptability, thereby determining the test material, and further completing the intelligent test question production of test knowledge points and test question types according to the design rules of different question types; finally, after manual review, the digital coding and storage of the test questions are realized. Among them, the test question production system is as follows: Figure 1 shown.

[0108] The specific implementation method includes the following three parts:

[0109] (1) Build a digital resource library to provide resources for test question production

[0110] Build a digital resource library including a knowledge point library, a question type feature library, and a material library, which are used for question writing knowledge point analysis, question type analysis, and question writing material acquisition when making questions. The specific resource library construction is as follows:

[0111] 1) Build a knowledge point library: The built knowledge point library contains knowledge points and their attribute information at each teaching stage, which is used for attribute analysis of the knowledge points in question making. Specifically, the knowledge point attributes are <knowledge point ID, subject, level stage, knowledge point category, mastery requirement, difficulty level, cognitive level, depth level>. Among them, the mastery requirement is divided into three levels: acquaintance, understanding, and application. The values are divided into f′, f″, f′′′ according to the three levels, and 0 < f′ < f″ < f′′′ < 1; the difficulty level is divided into three levels: easy, medium, and difficult. The values are divided into g′, g″, g′′′ according to the three levels, and 0 < g′ < g″ < g′′′ < 1; the cognitive level is determined by the average score of the teaching group answering the questions containing this knowledge point, and is divided into three levels: low, medium, and high. The values are divided into h′, h″, h′′′ according to the three levels, and 0 < h′ < h″ < h′′′ < 1; the depth level of the knowledge point refers to the depth of the three levels of basic layer, associated layer, and application layer divided according to the complexity and logical relationship of the knowledge point content in the knowledge system. The values are divided into k′, k″, k′′′ according to the three levels of depth, and 0 < k′ < k″ < k′′′ < 1.

[0112] 2) Question type feature library: The built question type feature library contains questions of various types, their attribute information, and question type feature information, which is used for attribute analysis of the question types in question making. Among them, the question attributes are <knowledge point, question type, difficulty, discrimination>. The question type feature information includes the question type features of various question types (such as multiple-choice, vocabulary filling, translation, etc.). Specifically as follows:

[0113]

[0114] Among them, the difficulty of the question is determined by the proportion of the number of people who answer the question correctly to the total number of people. The difficulty is assigned decimal numbers between 0 and 1 from easy to difficult; the discrimination of the question is obtained from the proportion of the number of people who answer the question correctly in the high-scoring group and the proportion of the number of people who answer the question correctly in the low-scoring group. In other words, the determination of the question difficulty is based on statistical principles, and the difficulty of the question is quantitatively evaluated by calculating the proportion between the number of people who answer the question correctly and the total number of people participating in the test. Specifically, the higher the correct rate, the easier the question, and the lower the correct rate, the more difficult the question. And this proportional relationship is transformed into a value between 0 and 1 for standardized representation, where approaching 0 means the question is more difficult, and approaching 1 means the question is easier; the calculation of the question discrimination is to divide the group participating in the test into a high-scoring group and a low-scoring group according to the overall performance, and then separately count the proportion of the number of people who answer the question correctly in the two groups. Subtract the correct rate of the low-scoring group from the correct rate of the high-scoring group to get the difference. This difference reflects the ability of the question to distinguish learners with different ability levels. The larger the difference, the more effectively the question can distinguish the true level differences of learners.

[0115] 3) Material library: Fragmented materials such as examples and model essays of various knowledge points are collected from the Internet, and the knowledge points in the materials are marked. The constructed material library is used to obtain materials for question creation when making test questions.

[0116] (2) Based on the designed AI test question creation engine, complete intelligent question editing

[0117] An AI test question production engine is designed. The engine first obtains the corresponding test question materials based on the test knowledge points, and then comprehensively evaluates the adaptability of the test materials from the dimensions of test content adaptability, test question type adaptability, test target adaptability, etc., so as to determine the test materials; further, according to the question type characteristics and question writing principles of the test type, the test rules are designed, and based on the determined test materials, the intelligent test question production of test knowledge points and question types is realized.

[0118] Step 1: Determine the material for the topic

[0119] like Figure 2 As shown, based on the determined question-editing knowledge points and question-editing question types, first obtain the question-editing materials corresponding to the question-editing knowledge points from the material library, and further intelligently analyze the matching between the attribute information such as the material content, material format, material difficulty, etc. of the question-editing materials and the attribute information such as the test question knowledge points, test question types, and test question difficulty, obtain the adaptability of the question-editing materials to the test content, test question types, and test objectives, and determine whether the question-editing materials are suitable for making test questions.

[0120] The specific implementation process is as follows:

[0121] Step 1: Determine the assessment weight of each knowledge point in the question-making material

[0122] Obtain the knowledge point information of the test material and assign assessment weights to the assessment knowledge points (i.e., the test knowledge points) and general knowledge points (i.e., the knowledge points other than the test knowledge points) in the test material. The assessment knowledge points are weighted as μ1, and the general knowledge points are weighted as μ2, with μ1 > μ2 > 0 (in this example, μ1 and μ2 are set to 0.8 and 0.2, respectively).

[0123] Step 2: Obtain the test content suitability results of the question-writing materials

[0124] Based on indicators such as the number of knowledge points included in the test material, the matching index between the assessment weight of each knowledge point in the test material and the mastery requirement, and the matching index between the assessment weight of each knowledge point in the test material and the cognitive level, the matching between the material content and the test knowledge points is analyzed, and the test content adaptability result f1 of the test material is obtained. The specific calculation is:

[0125]

[0126] Among them, f1 is the recommendation result of the test content adaptability of the question material; Min(a, b) is the total number of knowledge points contained in the question material, Max(a, b) is the knowledge points contained in the test knowledge point domain, min(a, b) is the average total number of knowledge points contained in this question type; Min(a, b)' is the number of test knowledge points in the question material, Min(a, b)'' is the number of general knowledge points in the question material, X i 、X j are the assessment weights of the i-th and j-th knowledge points in the test material; Y i 、Y j are the requirements for mastering the i-th and j-th knowledge points in the question-making materials; Match(X i , Y i )、Match(X j , Y j ) are the matching indexes of the assessment weights and mastery requirements of the i-th and j-th knowledge points in the test materials; Z i 、Z j are the cognitive levels of the teaching group for the i-th and j-th knowledge points in the question-making materials; Match(X i , Z i )、Match(X j , Z j ) are the matching indexes of the assessment weights of the i-th and j-th knowledge points in the test materials and the cognitive level of the teaching group; α1, α2, α3, δ1, δ2, ε1, and ε2 are the weight coefficients respectively.

[0127] Step 3: Obtain the test question type suitability results of the test material

[0128] Based on the matching index between the type of the editing material and the type of the editing question type, the matching index between the types of each knowledge point in the editing material and the type of the editing question type, and the matching index between the depth of each knowledge point in the editing material and the type of the editing question type, the matching of the material format and the test question type is analyzed, and the result f2 of the editing material's test question type adaptability is obtained. The specific calculation is:

[0129]

[0130] Among them, f2 is the result of the suitability of the test material to the test type; S is the type of the test material (text, chart, data, case, dialogue, etc.), S′ is the type of the test type (fill in the blank, multiple choice, translation, etc.); Match{(S), (S′)} is the matching index between the type of the test material and the type of the test type; P i 、P j For the i-th and j-th knowledge point types (words, phrases, syntax, common expressions, proper nouns, etc.) in the question-making material, match(Pi ,S′)、match(P j , S′) are the matching indexes of the i-th and j-th knowledge point types in the editing material and the types of the editing question types; Q i , Q j are the depth of the i-th and j-th knowledge points in the question-making materials; match(Q i , S′), Match(Q j , S′) are the matching indexes of the depth of the i-th and j-th knowledge points in the editing material and the type of the editing question; β1, β2, β3, θ1, θ2, are weight coefficients respectively.

[0131] Step 4: Obtain the test target suitability results of the editing materials

[0132] Based on indicators such as the difficulty of the material, the matching index between the difficulty of each knowledge point in the material and the difficulty of the question type, and the matching index between the difficulty of each knowledge point in the material and the cognitive level, the matching between the material difficulty and the test question difficulty is analyzed, and the test target adaptability result f3 of the material is obtained. The specific calculation is:

[0133]

[0134] Among them, f3 is the test target adaptation result of the editing material; R is the average difficulty of each knowledge point in the editing material, R′ is the average difficulty of the question type, Match{(R), (R′)} is the matching index between the difficulty of the editing material and the average difficulty of the question type; Min(a, b)′ is the number of test knowledge points in the editing material, Min(a, b)′′ is the number of general knowledge points in the editing material, R i 、R j are the difficulty levels of the i-th and j-th knowledge points in the question-making materials; Match(R i , R′), Match(R j , R′) are the matching indexes of the difficulty of the i-th and j-th knowledge points in the question-making materials and the difficulty of the question type; Z i 、Z j are the cognitive levels of the teaching group for the i-th and j-th knowledge points in the question-making materials; Match(R i , Z i )、Match(R j , Z j ) are the matching indexes of the difficulty of the i-th and j-th knowledge points in the question-making materials and the cognitive level of the teaching group; γ1, γ2, γ3, ρ1, ρ2, σ1, and σ2 are the weight coefficients respectively.

[0135] Step 5: Obtain the comprehensive adaptability results of the editing materials and determine the editing materials

[0136] ① Based on the fitness results of the above three dimensions and the weight distribution values ​​of each dimension, the comprehensive fitness result F of the editing material is obtained. The specific calculation is:

[0137] F=xf1+yf2+zf3

[0138] Among them, F is the comprehensive adaptability result of the question material; f1, f2, and f3 are the adaptability results of the examination content, examination question type, and examination target of the question material, respectively; x, y, and z are the weight values ​​of the three dimensions, respectively, and x+y+z=1.

[0139] ② According to the judgment result between the question material adaptation result F and the material adaptation threshold F′, determine whether the question material is suitable for the question knowledge point and question type to make the test question: if F is greater than F′, then it is determined that the question material is suitable; if F is less than F′, then the question material is not suitable.

[0140] Step 2: Intelligent question creation

[0141] Based on the characteristic information of each question type in the question type feature library, the question type characteristics of each question type are obtained. The characteristics of each question type cover multiple key dimensions, the most important of which include the question stem, answers and options. Then, according to the characteristics of each question type, question editing rules are designed, and based on the determined question editing materials, intelligent question creation of question editing knowledge points and question types is realized.

[0142] The rules for writing some question types are as follows:

[0143]

[0144] (3) Test questions stored in the database

[0145] After the test questions are manually reviewed, the system will digitize and encode them, and enter them into the test question type feature library to facilitate the subsequent application of the test questions in teacher teaching, student learning, etc.

[0146] The above embodiments have the following technical effects:

[0147] The above embodiment provides an artificial intelligence-based test question preparation method, which constructs digital resources such as a knowledge point library, a test question type feature library and a material library, and selects target knowledge points from the knowledge point library as question-writing knowledge points to obtain question-writing materials corresponding to the question-writing knowledge points.

[0148] The suitability of the test materials is assessed, including suitability for the test content, test question types, and test objectives. The test content suitability calculation incorporates factors such as the number of knowledge points contained in the test materials and the test knowledge point domain, the average number of knowledge points, the assessment weight of each knowledge point, mastery requirements, and the cognitive level of the teaching group. The test question type suitability calculation incorporates factors such as the type match between the test materials and the question type, the match between the type and depth of each knowledge point and the question type. The test objective suitability calculation incorporates factors such as the match between the average difficulty level of each knowledge point in the test materials and the question type, the difficulty level of each knowledge point, and the cognitive level of the teaching group. By weightedly calculating each suitability, the comprehensive suitability of the test materials is obtained, and candidate test materials whose suitability meets the threshold are screened out.

[0149] Furthermore, based on the question type feature information in the question type feature library, the design requirements for the question stem, answer, and options are obtained, and the question editing rules are designed accordingly. For example, for multiple-choice questions, key knowledge points are hollowed out from the example sentence materials through knowledge point recognition technology to form the question stem, and interference items are generated through the variations of knowledge points, synonyms and antonyms, collocations, etc., and randomly arranged with the correct answers to form options. For fill-in-the-blank questions, target knowledge points are extracted from the example sentence materials through knowledge point recognition technology, leaving them blank and retaining the first letters, and the extracted knowledge points are used as the answers. For translation questions, the example sentence materials and translations of the knowledge points are directly used as the question stem and answer, respectively.

[0150] Finally, test questions are automatically generated based on the candidate materials and the rules for question creation. After expert review and approval, standardized test questions are formed. Simultaneously, the generated test questions are digitally encoded and stored in a question type feature database, enabling standardized test question management.

[0151] The above embodiment constructs a refined attribute portrait from multiple dimensions such as the mastery requirements of knowledge points, degree of difficulty, cognitive level, and depth of levels, and evaluates the adaptability of the test materials from three dimensions: assessment content, question type characteristics, and target difficulty, thereby improving the matching degree between the test content and teaching requirements. At the same time, corresponding question editing rules are designed based on the characteristics of different question types, realizing the automatic generation and standardization of test questions. Compared with the existing technology, the above embodiment can generate high-quality test questions quickly, accurately, and on a large scale, reducing the burden of teacher preparation and improving the quality of test questions and teaching effectiveness.

[0152] The second embodiment of the present application relates to an artificial intelligence-based test question making system, the structure of which is as follows: Figure 4 As shown, the artificial intelligence-based test question making system includes:

[0153] A digital resource library construction module is used to construct a digital resource library including a knowledge point library, a test question type feature library, and a material library. The knowledge point library contains attribute information on the mastery requirements, difficulty level, cognitive level, and layer depth of the knowledge points. The test question type feature library contains question type feature information. The material library contains materials corresponding to the knowledge points.

[0154] A question writing material acquisition module is used to select target knowledge points from the knowledge point library as question writing knowledge points, determine the question writing type; based on the material library, obtain question writing materials corresponding to the question writing knowledge points;

[0155] The adaptability evaluation module is used to evaluate the adaptability of the editing material, including: calculating the adaptability of the examination content, the examination content adaptability represents the degree of matching between the content of the editing material and the editing knowledge points; calculating the adaptability of the examination question type, the examination question type adaptability represents the degree of matching between the format of the editing material and the editing question type; calculating the adaptability of the examination target, the examination target adaptability represents the degree of matching between the difficulty of the editing material and the difficulty of the test questions; and using a weighted calculation method to calculate the comprehensive adaptability of the editing material based on the examination content adaptability, the examination question type adaptability and the examination target adaptability;

[0156] A candidate material determination module is used to determine the editing materials with comprehensive adaptability higher than a set threshold as candidate editing materials;

[0157] A question editing rule design module is used to obtain question type features including question stems and answer generation requirements based on question type feature information in the question type feature library, and design question editing rules according to the question type features;

[0158] A test question generating module, configured to generate test questions based on the candidate test question materials and the test question editing rules;

[0159] The test question storage module is used to digitally encode the generated test questions after manual review and store them in the test question type feature library to complete the test question storage.

[0160] The first embodiment is a method embodiment corresponding to the present embodiment. The technical details in the first embodiment can be applied to the present embodiment, and the technical details in the present embodiment can also be applied to the first embodiment.

[0161] It should be noted that those skilled in the art should understand that the implementation functions of the modules shown in the embodiment of the above-mentioned artificial intelligence-based test question making system can be understood with reference to the relevant description of the aforementioned artificial intelligence-based test question making method. The functions of the modules shown in the embodiment of the above-mentioned artificial intelligence-based test question making system can be implemented by a program (executable instruction) running on a processor, or by a specific logic circuit. If the artificial intelligence-based test question making system of the present application embodiment is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk. In this way, the embodiment of the present application is not limited to any specific combination of hardware and software.

[0162] Accordingly, an embodiment of the present application further provides a computer storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, the various method embodiments of the present application are implemented.

[0163] In addition, the embodiment of the present application also provides an artificial intelligence-based test question making system, which includes a memory for storing computer-executable instructions, and a processor; the processor is used to implement the steps in the above-mentioned method implementation methods when executing the computer-executable instructions in the memory. Among them, the processor can be a central processing unit (Central Processing Unit, referred to as "CPU"), or other general-purpose processors, digital signal processors (Digital Signal Processor, referred to as "DSP"), application-specific integrated circuits (Application Specific Integrated Circuit, referred to as "ASIC"), etc. The aforementioned memory can be a read-only memory (read-only memory, referred to as "ROM"), a random access memory (random access memory, referred to as "RAM"), a flash memory (Flash), a hard disk or a solid-state drive, etc. The steps of the method disclosed in each embodiment of the present application can be directly embodied as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0164] It should be noted that in this patent application, relational terms such as first and second, etc., are used solely to distinguish one entity or operation from another, and do not necessarily require or imply any actual relationship or order between these entities or operations. Furthermore, the terms "comprise," "include," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. Without further limitation, an element specified by the phrase "comprising a" does not preclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element. In this patent application, reference to performing an action in accordance with an element means performing the action in accordance with at least that element, including two situations: performing the action in accordance with that element alone, and performing the action in accordance with that element and other elements. Expressions such as "plurality," "multiple times," and "many" include "two," "twice," "two kinds," and "more than two," "more than two times," and "more than two kinds."

[0165] All documents mentioned in this application are considered to be included in their entirety in the disclosure of this application so that they can be used as a basis for modification when necessary. In addition, it should be understood that after reading the above disclosure of this application, those skilled in the art may make various changes or modifications to this application, and these equivalent forms also fall within the scope of protection claimed in this application.

Claims

1. A test question making method based on artificial intelligence, characterized in that: It includes the following steps: Construct a digital resource library including a knowledge point library, a test question type feature library, and a material library. The knowledge point library contains the mastery requirements, difficulty levels, cognitive levels, and hierarchical depth attribute information of knowledge points. The test question type feature library contains question type feature information. The material library contains materials corresponding to knowledge points; Select a target knowledge point from the knowledge point library as the knowledge point for question compilation and determine the question compilation type; Based on the material library, obtain the question compilation materials corresponding to the knowledge point for question compilation; Conduct an adaptability assessment on the question compilation materials, including: calculating the adaptability of the inspected content, which represents the matching degree between the content of the question compilation materials and the knowledge point for question compilation; calculating the adaptability of the inspected question type, which represents the matching degree between the format of the question compilation materials and the question compilation type; calculating the adaptability of the inspected target, which represents the matching degree between the difficulty of the question compilation materials and the difficulty of the test questions; and calculating the comprehensive adaptability of the question compilation materials based on the adaptability of the inspected content, the adaptability of the inspected question type, and the adaptability of the inspected target using a weighted calculation method; Determine the question compilation materials with a comprehensive adaptability higher than the set threshold as candidate question compilation materials; Based on the question type feature information in the test question type feature library, obtain question type features including the question stem and answer generation requirements, and design question compilation rules according to the question type features; Generate test questions based on the candidate question compilation materials and the question compilation rules; Digitally encode the generated test questions after manual review and store them in the test question type feature library to complete the warehousing of test questions.

2. The method according to claim 1, wherein The mastery requirements are divided into 3 levels: recognition and understanding, comprehension and explanation, and mastery and application, with values f′, f″, f′′′ respectively, and 0 < f′ < f″ < f′′′ < 1; the difficulty levels are divided into 3 levels: easy, medium, and difficult, with values g′, g″, g′′′ respectively, and 0 < g′ < g″ < g′′′ < 1; the cognitive level is determined by the average score of the teaching group answering test questions containing this knowledge point, and is divided into 3 levels: low, medium, and high, with values h′, h″, h′′′ respectively, and 0 < h′ < h″ < h′′′ < 1; the hierarchical depth refers to the depth of 3 levels: basic layer, associated layer, and application layer, divided according to the complexity and logical relationship of the knowledge point content in the knowledge system, with values k′, k″, k′′′ respectively, and 0 < k′ < k″ < k′′′ < 1.

3. The method according to claim 1, wherein The adaptability of the inspected content is calculated by the following formula: Among them, Match(A,B) represents the matching degree between A and B; Min(a,b) is the total number of knowledge points contained in the question material, Max(a,b) is the total number of knowledge points contained in the test knowledge point domain, and min(a,b) is the average total number of knowledge points contained in this question type; X i 、X j are the assessment weights of the i-th and j-th knowledge points in the test material; Y i 、Y j are the requirements for mastering the i-th and j-th knowledge points in the question-making materials; Z i , Z j are the cognitive levels of the teaching group for the i-th and j-th knowledge points in the question-making materials; α1, α2, α3, δ1, δ2, ε1, and ε2 are the weight coefficients respectively.

4. The method according to claim 1, wherein The adaptability of the inspected question type is calculated by the following formula: Among them, Match(A,B) represents the matching degree between A and B; S is the type of the editing material, S′ is the type of the editing question type; P i 、P j is the type of knowledge point i and j in the question-making material; Q i , Q j are the depths of the i-th and j-th knowledge points in the question-making materials; β1, β2, β3, θ1, θ2, are weight coefficients respectively.

5. The method according to claim 1, wherein The adaptability of the inspected target is calculated by the following formula: Among them, Match(A,B) represents the matching degree between A and B; R is the average difficulty of each knowledge point in the question material, R′ is the average difficulty of the question type; R i 、R j are the difficulty levels of the i-th and j-th knowledge points in the question-making materials; Z i , Z j are the cognitive levels of the teaching group for the i-th and j-th knowledge points in the question-making materials; γ1, γ2, γ3, ρ1, ρ2, σ1, and σ2 are the weight coefficients respectively.

6. The method according to claim 1, wherein The calculation formula for the comprehensive adaptability is: F = xf1 + yf2 + zf3 where F is the result of the comprehensive adaptability of the question compilation materials, f1, f2, and f3 are the adaptability of the inspected content, the adaptability of the inspected question type, and the adaptability of the inspected target respectively, x, y, and z are the weight values of 3 dimensions respectively, and 0 < x < 1, 0 < y < 1, 0 < z < 1, x + y + z = 1.

7. The method according to claim 1, wherein For multiple-choice questions, the question-writing rules include: based on the example sentence materials of the question-writing knowledge points, the target knowledge points are hollowed out through knowledge point recognition technology to generate the question stem; interference items are generated in priority order through the changed word forms, synonyms and antonyms, and phrase collocations of the target knowledge points; the target knowledge points and interference items that are hollowed out are randomly sorted to generate option groups and answers.

8. The method according to claim 1, wherein For vocabulary fill-in-the-blank questions, the question-writing rules include: extracting target knowledge points from example sentence materials based on the question-writing knowledge points through knowledge point recognition technology, adding the first letter prompt information of the knowledge point while leaving blanks to generate the question stem; and directly using the extracted target knowledge point as the answer to the question.

9. The method according to claim 1, wherein For translation questions, the question-making rules include: using example sentences of the question-making knowledge points as the question stem; and using the translation sentences corresponding to the example sentences as the answers to the questions.

10. An artificial intelligence-based test question making system, characterized in that: include: A digital resource library construction module is used to construct a digital resource library including a knowledge point library, a test question type feature library, and a material library. The knowledge point library contains attribute information on the mastery requirements, difficulty level, cognitive level, and layer depth of the knowledge points. The test question type feature library contains question type feature information. The material library contains materials corresponding to the knowledge points. A question-writing material acquisition module is used to select target knowledge points from the knowledge point library as question-writing knowledge points and determine the question type; Based on the material library, obtaining the question-writing materials corresponding to the question-writing knowledge points; The adaptability evaluation module is used to evaluate the adaptability of the editing material, including: calculating the adaptability of the examination content, the examination content adaptability represents the degree of matching between the content of the editing material and the editing knowledge points; calculating the adaptability of the examination question type, the examination question type adaptability represents the degree of matching between the format of the editing material and the editing question type; calculating the adaptability of the examination target, the examination target adaptability represents the degree of matching between the difficulty of the editing material and the difficulty of the test questions; and using a weighted calculation method to calculate the comprehensive adaptability of the editing material based on the examination content adaptability, the examination question type adaptability and the examination target adaptability; A candidate material determination module is used to determine the editing materials with comprehensive adaptability higher than a set threshold as candidate editing materials; A question editing rule design module is used to obtain question type features including question stems and answer generation requirements based on question type feature information in the question type feature library, and design question editing rules according to the question type features; A test question generating module, configured to generate test questions based on the candidate test question materials and the test question editing rules; The test question storage module is used to digitally encode the generated test questions after manual review and store them in the test question type feature library to complete the test question storage.

Images