Evaluation question dynamic generation method and system
Through the dynamic generation of evaluation question system, AST analysis and combination technology is used to solve the problem of problem generation in online education platforms and mathematical formula processing, and cross-platform adaptive evaluation question generation is achieved.
Patent Information
- Application Number
- CN202510361467.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-26
- Publication Date
- 2025-07-18
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing online education platform's evaluation question generation system is not flexible enough, and it is impossible to dynamically adjust the difficulty or structure of the question, and the mathematical formula processing ability is weak, the generation logic is single, and it is difficult to combine multiple knowledge points or dynamic parameters.
Complex questions are generated through input parameter configuration, formula template matching and parsing into an abstract syntax tree (AST), combined with dynamic parameter replacement and logical generation, and complex questions are generated, and cross-platform rendering is supported through difficulty verification and multimodal output.
It realizes in-depth and flexible generation of the question structure, avoids difficulty mismatch, improves the randomness and accuracy of the evaluation, and supports cross-platform rendering of mobile, web pages and paper test papers.
Smart Images
Figure CN120337903A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of online education, and particularly relates to a method and system for dynamically generating assessment questions. Background Art
[0002] Course recommendation on an online education platform refers to an intelligent service that accurately matches courses that meet the learning needs, interests, and abilities of students through technical means to improve learning efficiency and user experience. The core goals of course recommendation include: personalized matching: recommending customized courses based on students' learning behaviors, interest preferences, knowledge levels, etc.; increasing participation: reducing the course selection cost of students through accurate recommendation and increasing the course completion rate; optimizing resource allocation: balancing the exposure of popular courses and long-tail courses to improve the resource utilization rate of the platform.
[0003] In order to verify the learning level of students, it is necessary to conduct tests through assessment questions. However, most existing question generation systems rely on static question banks and have the following defects: lack of flexibility: unable to dynamically adjust the difficulty or structure of questions according to the user's level; weak ability to process mathematical formulas: most systems only support text questions and lack support for parsing, generating, and rendering complex formulas; single generation logic: the question generation rules are fixed, and it is difficult to combine multiple knowledge points or dynamic parameters. Summary of the Invention
[0004] The technical problem to be solved by the present invention is to overcome the above-mentioned defects of the prior art and provide a method and system for dynamically generating assessment questions.
[0005] The technical solution adopted to solve the above technical problem is: a method for dynamically generating assessment questions, including the following specific steps:
[0006] Step 1: Input parameter configuration, receiving the target parameters input by the user, including knowledge points, difficulty level, and question type;
[0007] Step 2: Formula template matching and parsing, matching the mathematical formula structure corresponding to the target knowledge point from the formula template library and parsing it into an Abstract Syntax Tree (AST);
[0008] Step 3: Dynamic parameter substitution and logic generation, replacing the variables in the template according to the difficulty, and combining multiple formula templates to generate complex questions;
[0009] Step 4: Difficulty verification and optimization, verifying whether the generated questions meet the target difficulty through a pre-trained difficulty prediction model, and if not, readjusting the parameters;
[0010] Step 5: Multi-modal question output, rendering the questions into LaTeX, images, and draggable geometric figures.
[0011] Through the above technical solutions, the deep and flexible generation of the question structure can be achieved. By combining parameter randomization, structural complexity, and historical data to dynamically adapt the difficulty, the questions can be output in formats suitable for mobile devices, web pages, or paper test papers, supporting cross-platform rendering.
[0012] Furthermore, the difficulty levels are divided into L1 - L5, and the question types include calculation, proof, application, and statement.
[0013] Through the above technical solutions, the situation where the difficulty of the generated questions does not match can be avoided, thus preventing it from affecting the assessment.
[0014] Furthermore, the Abstract Syntax Tree (AST) of the mathematical formula represents operators, operation trees, and logical relationships. The node types include: operator nodes, variable nodes, numerical nodes, function nodes, and relational nodes.
[0015] Through the above technical solutions, the mathematical formula can be abstracted, thereby decomposing each node in the mathematical formula.
[0016] Furthermore, the variable nodes include symbolic variables and dynamic parameters, and the numerical nodes are randomly generated numerical values.
[0017] Through the above technical solutions, the randomness of the generated assessment questions can be improved, avoiding the occurrence of duplicate questions.
[0018] Furthermore, the dynamic operations of the AST and question generation include the following specific steps:
[0019] Step 1: Variable substitution and parameterization. Replace the variable nodes in the AST with random values, and then replace the entire subtree to change the formula structure.
[0020] Step 2: Structure combination and multi - knowledge - point integration. Combine multiple ASTs to generate complex questions.
[0021] Step 3: Constraint and verification of the AST. To ensure that the generated questions are correct, apply constraint rules in the AST, including mathematical constraints and difficulty constraints.
[0022] Through the above technical solutions, it is ensured that the generated questions can be answered and meet the corresponding difficulty level.
[0023] Furthermore, it includes a user interface module, a formula parsing engine, a logic generation engine, a difficulty adaptation module, and a question bank database. The user interface module is responsible for configuring parameters and displaying questions. The formula parsing engine is responsible for constructing a formula syntax parser based on ANTLR, supporting formula structure splitting and semantic analysis.
[0024] Further, the logic generation engine is responsible for combining a random algorithm with a constraint solver to ensure the correctness of the generated questions. The difficulty adaptation module is responsible for recommending questions of similar difficulty using a collaborative filtering algorithm and adjusting the parameter range through a decision tree. The question bank database is responsible for storing formula templates, historical questions, and user answer records.
[0025] The beneficial effects of the present invention are as follows: The present invention realizes the deep and flexible generation of the question structure through AST parsing and combination, dynamically adapts the difficulty by combining parameter randomization, structural complexity, and historical data. The questions can be output in a format suitable for mobile devices, web pages, or paper test papers, supporting cross-platform rendering. Description of the Drawings
[0026] Figure 1 is the flowchart of the method of the present invention. Detailed Embodiments
[0027] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0028] As Figure 1 shown, the method and system for dynamically generating assessment questions in this embodiment include the following specific steps:
[0029] Step 1: Input parameter configuration, receiving the target parameters input by the user, including knowledge points, difficulty level, and question type;
[0030] Step 2: Formula template matching and parsing, matching the mathematical formula structure corresponding to the target knowledge point from the formula template library and parsing it into an abstract syntax tree (AST);
[0031] Step 3: Dynamic parameter substitution and logic generation, replacing the variables in the template according to the difficulty, and combining multiple formula templates to generate complex questions;
[0032] Step 4: Difficulty verification and optimization, verifying whether the generated questions meet the target difficulty through a pre-trained difficulty prediction model, and if not, readjusting the parameters;
[0033] Step 5: Multi-modal question output, rendering the questions into LaTeX, images, and draggable geometric figures, which can realize the deep and flexible generation of the question structure, dynamically adapt the difficulty by combining parameter randomization, structural complexity, and historical data. The questions can be output in a format suitable for mobile devices, web pages, or paper test papers, supporting cross-platform rendering.
[0034] The difficulty level is divided into L1-L5, and the question types include calculation, proof, application, and statement, which can avoid the mismatch of the difficulty of the generated questions and affect the assessment.
[0035] The Abstract Syntax Tree (AST) of the mathematical formula represents operators, operation trees, and logical relationships. The node types include: operator nodes, variable nodes, numerical nodes, function nodes, and relationship nodes. The mathematical formula can be abstracted, thereby decomposing each node in the mathematical formula.
[0036] The variable nodes include symbolic variables and dynamic parameters, and the numerical nodes are randomly generated numerical values, which can improve the randomness of generating assessment questions and avoid the appearance of duplicate questions.
[0037] The dynamic operations and question generation of the AST include the following specific steps:
[0038] Step 1: Variable substitution and parameterization. Replace the variable nodes in the AST with random values, and then replace the entire subtree to change the formula structure.
[0039] Step 2: Structure combination and multi-knowledge point integration. Combine multiple ASTs to generate complex questions.
[0040] Step 3: Constraint and verification of the AST. To ensure the correctness of the generated questions, apply constraint rules in the AST, including mathematical constraints and difficulty constraints, to ensure that the generated questions can be answered and meet the corresponding difficulty level.
[0041] It includes a user interface module, a formula parsing engine, a logic generation engine, a difficulty adaptation module, and a question bank database. The user interface module is responsible for configuring parameters and presenting questions. The formula parsing engine is responsible for building a formula syntax parser based on ANTLR, supporting formula structure splitting and semantic analysis.
[0042] The logic generation engine is responsible for combining a random algorithm and a constraint solver to ensure the correctness of the generated questions. The difficulty adaptation module is responsible for recommending questions of similar difficulty using a collaborative filtering algorithm and adjusting the parameter range through a decision tree. The question bank database is responsible for storing formula templates, historical questions, and user answer records.
[0043] The above is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention.
Claims
1. A method for dynamically generating assessment questions, characterized in that It includes the following specific steps: Step 1: Input parameter configuration, receiving the target parameters input by the user, including knowledge points, difficulty levels, and question types; Step 2: Formula template matching and parsing, matching the mathematical formula structure corresponding to the target knowledge point from the formula template library and parsing it into an Abstract Syntax Tree (AST); Step 3: Dynamic parameter substitution and logic generation, replacing the variables in the template according to the difficulty, and combining multiple formula templates to generate complex questions; Step 4: Difficulty verification and optimization, verifying whether the generated questions meet the target difficulty through a pre-trained difficulty prediction model, and adjusting the parameters again if not satisfied; Step 5: Multi-modal question output, rendering the questions into LaTeX, images, and draggable geometric figures.
2. The method for dynamically generating assessment questions according to claim 1, wherein The difficulty levels are divided into L1 - L5, and the question types include calculation, proof, application, and statement.
3. The method for dynamically generating assessment questions according to claim 2, wherein The Abstract Syntax Tree (AST) of the mathematical formula represents operators, operation trees, and logical relationships, and the node types include: operator nodes, variable nodes, numerical nodes, function nodes, and relationship nodes.
4. The method for dynamically generating evaluation questions according to claim 3, wherein The variable nodes include symbolic variables and dynamic parameters, and the numerical nodes are randomly generated numerical values.
5. The method for dynamically generating evaluation questions according to claim 4, characterized in that, The dynamic operations of the AST and question generation include the following specific steps: Step 1: Variable substitution and parameterization, replacing the variable nodes in the AST with random values, and then replacing the entire subtree to change the formula structure; Step 2: Structure combination and multi-knowledge point integration, combining multiple ASTs to generate complex questions; Step 3: Constraint and verification of the AST, applying constraint rules in the AST to ensure the correctness of the generated questions, including mathematical constraints and difficulty constraints.
6. The dynamic generation system for assessment questions according to claim 5, wherein It includes a user interface module, a formula parsing engine, a logic generation engine, a difficulty adaptation module, and a question bank database. The user interface module is responsible for configuring parameters and displaying questions. The formula parsing engine is responsible for building a formula syntax parser based on ANTLR, supporting formula structure splitting and semantic analysis.
7. The dynamic generation system for evaluation questions according to claim 6, wherein The logic generation engine is responsible for combining random algorithms and constraint solvers to ensure the correctness of the generated questions. The difficulty adaptation module is responsible for recommending similar difficulty questions using collaborative filtering algorithms and adjusting the parameter range through decision trees. The question bank database is responsible for storing formula templates, historical questions, and user answer records.