Dynamically-configured accounting document question setting and test paper composition method and dynamically-configured accounting document question setting and test paper composition system

Abstract

The invention discloses a dynamically configured accounting document question setting and test paper composition method and system, relates to the field of accounting education, and provides the following scheme: performing feature extraction and comparison on question solving information and solution sets based on a machine learning technology to generate used and unused solution sets; screening the effective solutions which are not fully used by using a threshold value, and counting the occurrence times of the effective solutions in the historical solution set as effective rates for sorting; learning grades are divided according to historical scores of students, and targeted questions are generated by differentially retaining an efficient solution; and constructing a test question template based on the question category, and performing dynamic test paper composition by utilizing a machine learning optimization algorithm. Deep mining of hidden knowledge points is realized through intelligent analysis of a solution using mode; through accurate assessment of student ability and question adaptation, the problems of simplification and staticization of a traditional test paper composition method are solved, and the personalized level and training effect of accounting document teaching are improved.

Description

Technical Field

[0001] This invention relates to the field of accounting education technology, and more specifically, this application relates to a dynamically configurable method and system for generating and compiling accounting questions. Background Technology

[0002] With the continuous development of accounting education and vocational training, the application of accounting document processing skills in financial practice and auditing operations is becoming increasingly important. Accounting document exam questions are a core component of accounting training, covering key aspects such as voucher preparation, ledger entry, and financial statement compilation. If the question-setting and test-compilation process is not scientific and personalized, it may lead to poor learning outcomes and weak skill mastery among students, impacting their practical work abilities. Therefore, optimizing the methods for setting and compiling accounting document exam questions has become an issue to be addressed in accounting education.

[0003] Existing accounting document question generation solutions mostly rely on teachers' experience or fixed question banks for manual question generation. The approach involves manually selecting questions from a predefined question bank and rules, based on knowledge points and difficulty levels, to create an exam paper. The advantage of this technology is its simplicity, low cost, and ability to quickly generate basic exam papers that meet standard teaching needs.

[0004] However, existing question-generating methods are easily influenced by subjective factors during question selection and lack data-driven optimization, resulting in mismatches between question difficulty and student level, repetitive solutions, or incomplete coverage, leading to poor test paper generation effectiveness. Regarding dynamically adapting to student needs, existing technologies often fail to effectively utilize historical solution data and cannot personalize adjustments based on individual student differences. This manifests as monotonous and untargeted question solutions during test paper generation, easily leading to low student practice efficiency and insufficient learning interest. Therefore, this paper proposes a dynamically configurable accounting document question-generating method and system to address this problem. Summary of the Invention

[0005] To address the aforementioned technical problems, this technical solution resolves the issues raised in the background section.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] Firstly, this application provides a dynamically configurable method for generating and compiling exam questions based on accounting documents, including:

[0008] Obtain the sequence of solution information from accounting document test questions and students' historical accounting document solution sets. The solution information includes question category, question stem information, solution information, and a set of solutions associated with question category and question stem information.

[0009] Machine learning is used to extract features from the problem-solving information and the solutions in the solution set to generate common features and distinguishable features. The common features are those shared by both, and the distinguishable features are those unique to the problem-solving information.

[0010] The solutions extracted from the problem-solving information are compared with the corresponding solution sets in terms of common features and distinguishing features to generate a set of used solutions and a set of unused solutions.

[0011] The usage value of each solution in the set of used solutions is counted. If the usage value is lower than the preset usage threshold, the solution is added to the set of unused solutions.

[0012] Count the number of times each solution in the unused solution set appears in the historical accounting document solution set, and use the number of occurrences as the efficiency. Sort the unused solution set in descending order according to the efficiency.

[0013] Students are classified into learning levels based on the average score of their historical accounting document problem-solving set within a specified time window, and a predetermined number of unused solutions from the solution set are retained in descending order based on the learning level.

[0014] The accounting document test questions are used to generate targeted questions based on the retained set of unused solutions. These targeted questions are then grouped by category to obtain group document test questions. Finally, group test question templates are obtained by combining the knowledge points and difficulty levels of the group document test questions.

[0015] The test question templates for the aforementioned groups are dynamically configured, and machine learning optimization algorithms are used to process the test papers, resulting in accounting document test papers.

[0016] Secondly, this application provides a dynamically configurable accounting document question generation and test paper preparation system for implementing the dynamically configurable accounting document question generation and test paper preparation method described in any of the above claims, including:

[0017] The data acquisition module is used to acquire the sequence of solution information from accounting document test questions and students' historical accounting document solution sets. The solution information includes question category, question stem information, solution information, and a set of solutions associated with question category and question stem information.

[0018] The feature extraction module is used to extract features from the problem-solving information and the solutions in the solution set using machine learning, generating common features and distinguishing features. The common features are those shared by both, and the distinguishing features are the unique features of the problem-solving information.

[0019] The feature comparison module is used to compare the common features and distinguishing features of the solutions extracted from the problem-solving information with the corresponding set of solutions, and to generate a set of used solutions and a set of unused solutions.

[0020] The solution set update module is used to count the usage value of each solution in the used solution set. If the usage value is lower than the preset usage threshold, the solution is added to the unused solution set.

[0021] The solution set descending order module is used to count the number of times each solution in the unused solution set appears in the historical accounting document solution set, and uses the number of occurrences as the effectiveness rate to sort the unused solution set in descending order according to the effectiveness rate;

[0022] The solution set retention module is used to classify students’ learning levels based on the average score of their historical accounting document solution set within a specified time window, and to retain a predetermined number of unused solutions in the solution set in descending order based on the learning level.

[0023] The test question template acquisition module is used to generate targeted questions from the accounting document test questions based on the retained set of unused solutions, group the targeted questions by category to obtain group document test questions, and combine them based on the knowledge points and difficulty levels of the group document test questions to obtain group test question templates;

[0024] The test paper generation module is used to dynamically configure the test question templates for the groups and use machine learning optimization algorithms to generate test papers for accounting documents.

[0025] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described dynamically configured method for generating and assembling accounting document questions.

[0026] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for dynamically configured accounting document question generation and test paper preparation.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] 1. This application uses machine learning technology to extract and compare features of problem-solving information and solution sets, generates a set of used and unused solutions, and filters out effective solutions that are not fully used based on the usage value threshold. This solves the problem of single and repetitive problem-solving methods in traditional test paper generation methods, and realizes in-depth mining and utilization of hidden knowledge points.

[0029] 2. This application divides learning levels based on the average score of students' historical problem-solving data within a time window, and accordingly retains efficient solutions from the set of unused solutions. This solves the problem that traditional methods cannot adapt to individual student differences, achieves precise adaptation of personalized question generation, and effectively improves the pertinence of learning and training. Attached Figure Description

[0030] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Wherein:

[0031] Figure 1 This is a flowchart of the dynamically configured accounting document question generation and test paper assembly method proposed in this invention;

[0032] Figure 2 This is a flowchart of the method for validating the set of unused solutions in this invention;

[0033] Figure 3 This is a data flow diagram of the document assembly method in this invention;

[0034] Figure 4 This is a structural block diagram of the dynamically configurable accounting document question generation and test paper system proposed in this invention. Detailed Implementation

[0035] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0036] In existing technologies, accounting document-based question generation suffers from static question configuration and a lack of personalized adaptation and dynamic optimization capabilities. It typically employs a fixed question bank combination; if the question difficulty doesn't match the students' level or the solution methods are incomplete, the system cannot automatically extract knowledge points or optimize the questions, still heavily relying on teacher experience for adjustments. This static and generic nature not only reduces question generation efficiency but may also lead to poor student learning outcomes and weak skill mastery due to the lack of targeted reinforcement training.

[0037] To address the aforementioned issues, this application utilizes machine learning techniques to extract common and distinctive features from problem-solving information and solution sets, generating sets of used and unused solutions. Low-frequency solutions are filtered based on usage thresholds, and solution effectiveness is calculated by combining historical occurrence counts, constructing an optimized solution sequence. Unused solutions are retained according to student learning level differences, generating targeted questions and constructing test templates. Finally, machine learning optimization algorithms dynamically configure test paper generation. Through in-depth mining of problem-solving features and dynamic analysis of solution usage patterns, this application achieves precise adaptation of question difficulty and optimized balance of knowledge point coverage, solving the problems of the singularity and static nature of traditional test paper generation methods. This ensures a gradual match between training content and student abilities, improving the effectiveness and efficiency of accounting document teaching.

[0038] Example 1

[0039] like Figure 1 As shown, a method for dynamically configurable accounting document question generation and test paper preparation is introduced, including:

[0040] S1. Obtain the sequence of solution information from accounting document test questions and students' historical accounting document solution sets. The solution information includes question type, question stem information, solution information, and a set of solutions related to question type and question stem information.

[0041] S2. Use machine learning to extract features from the problem-solving information and the solutions in the solution set, generating common features and distinctive features. Common features are those shared by both, while distinctive features are the unique features of the problem-solving information.

[0042] S3. Compare the common features and distinguishing features of the solutions extracted from the problem-solving information with the corresponding solution set to generate a set of used solutions and a set of unused solutions;

[0043] S4. Calculate the usage value of each solution in the set of used solutions. If the usage value is lower than the preset usage threshold, then add the solution to the set of unused solutions.

[0044] S5. Count the number of times each solution in the unused solution set appears in the historical accounting document solution set, and use the number of occurrences as the effectiveness rate. Sort the unused solution set in descending order according to the effectiveness rate.

[0045] S6. Based on the average score of the students' historical accounting document problem-solving set within the specified time window, divide the students' learning levels, and retain a predetermined number of unused solutions in the solution set after descending order based on the learning level.

[0046] S7. Generate targeted questions from the accounting document test questions based on the retained set of unused solutions, group the targeted questions by category to obtain group document test questions, and combine the knowledge points and difficulty levels of the group document test questions to obtain group test question templates.

[0047] S8. Dynamically configure the group test question templates and use machine learning optimization algorithms to process the test papers to obtain accounting document test papers.

[0048] Regarding step S1:

[0049] Raw data was collected from a pre-set accounting document question bank and a collection of students' historical accounting document problem-solving experiences. Specifically:

[0050] Accounting document question bank: Stored in a database or storage medium, it contains multiple accounting document questions. Each question is associated with a question category (such as voucher preparation, ledger entry, report preparation, etc.), question stem information (such as question description, document format, business scenario), and a corresponding set of solutions (i.e., multiple feasible solutions related to the question category and question stem information).

[0051] Student Historical Accounting Document Problem Solving Collection: Records students' problem-solving records for accounting document test questions within a historical time window. Each problem-solving record includes the question category, question stem information, problem-solving information (such as the actual problem-solving steps used by the student, the answer, and the score), and the associated set of solutions (i.e. the standard or common solutions corresponding to the question).

[0052] The acquired raw data is parsed and serialized to construct a sequence of problem-solving information:

[0053] Data cleaning and formatting were performed on accounting document test questions and students' historical solution sets to ensure the integrity and consistency of fields such as question category, question stem information, solution information, and solution set.

[0054] Each problem-solving instance (i.e., a test question and its corresponding solution record) is encapsulated as a problem-solving information unit. Each unit includes: Question Category: Identifying the knowledge area to which the test question belongs, such as "transfer voucher preparation" or "profit and loss statement preparation"; Question Stem Information: Describing the specific content of the test question, including text description, document template, or numerical parameters; Solution Information: Recording key data in the student's problem-solving process, such as the steps used, calculation results, error markers, or timestamps; Solution Set: Multiple solutions associated with the question category and question stem information. Each solution represents a feasible problem-solving method or strategy (such as "direct method", "indirect method", or "journal entry method").

[0055] Multiple problem-solving information units are arranged in chronological or problem-category order to generate a problem-solving information sequence. This sequence is an ordered list used for subsequent feature extraction and comparison operations.

[0056] Regarding step S2:

[0057] For example, the process of using machine learning to extract features from problem-solving information and solutions in a set of solutions, and generating common and distinctive features, includes:

[0058] Obtain the text description and document format template of the question stem;

[0059] Extract keywords, semantic vectors, and syntactic structures from the question stem using natural language processing techniques;

[0060] Encode each solution in the solution set and extract its method type, applicable conditions, computational logic, and step sequence characteristics;

[0061] The similarity between the solution features of the problem-solving information and the solution features of the solution set is calculated and pattern matched.

[0062] Features that exist in both the problem-solving information and the solution set are taken as common features, while features that are unique to the problem-solving information but not included in the solution set are taken as distinguishing features.

[0063] Specifically, the data obtained in step S1 includes: Question category: such as "voucher preparation" or "ledger registration"; Question stem information: including the question text description, document format template, or business scenario parameters; Solution information: recording the steps, calculation results, or error markers in the student's actual problem-solving process; and Solution set: multiple standard or common solutions associated with the question category and question stem information.

[0064] Machine learning techniques are used to extract and classify solution-related features from the acquired data, including:

[0065] (1) Feature extraction

[0066] Natural Language Processing (NLP) techniques are used to parse the text content in the question stem and solution information, extracting keywords, semantic vectors, and syntactic structures.

[0067] Encode each solution in the solution set and extract its characteristics such as method type (e.g., "direct method", "indirect method", "journal entry method"), applicable conditions, calculation logic and step sequence.

[0068] Based on numerical analysis methods, quantitative features are extracted from numerical parameters (such as amount, date, account number) and calculation process data in the problem-solving information.

[0069] (2) Feature comparison and classification

[0070] The similarity calculation and pattern matching are performed between the actual solution features used in the problem-solving information and the standard solution features in the solution set.

[0071] Common features: Identify common features found in the problem-solving information and solution set, such as shared keywords, identical calculation steps, consistent method types, or shared document elements. For example, in the "transfer voucher preparation" problem, the "debit and credit balance" principle is a common feature between the two.

[0072] Distinguishing features: Identify features unique to the problem-solving information that are not included in the set of solutions, such as unique errors made by students, unconventional calculation paths, or personalized annotations. For example, if a student uses a non-standard account classification method during the problem-solving process, this is a unique feature of the problem-solving information.

[0073] (3) Feature representation

[0074] The extracted common features and distinguishing features are represented as feature vectors or structured feature sets, respectively, to facilitate comparison and use by subsequent modules.

[0075] Output the following results for subsequent processing: Common feature set: contains feature data that exists in both the problem-solving information and the solution set, represented in a structured form; Distinguishing feature set: contains feature data that is unique to the problem-solving information but does not appear in the solution set, also represented in a structured form.

[0076] Regarding step S3:

[0077] Obtain the data in step S2: Solution features extracted from problem-solving information: This set contains unique features extracted from students' actual problem-solving information, representing the characteristics of the problem-solving methods actually used by students; Common features of the solution set: This set contains features common to the problem-solving information in the standard solution set corresponding to the problem.

[0078] All of the above data are stored in the form of structured feature sets or feature vectors.

[0079] The input feature data is systematically compared and classified. The specific process includes:

[0080] (1) Feature comparison

[0081] The extracted solution features are compared with the common features of the solution set using similarity matching and consistency comparison. The comparison process uses feature vector distance calculation or pattern matching algorithms to evaluate the degree of correlation between the actual solution features and the standard solution features.

[0082] For each solution, analyze its usage in the actual problem-solving process: if the features extracted from the problem-solving information highly match the common features of a standard solution, then the solution is considered to have been used; if the features of a standard solution do not appear in the problem-solving information, then the solution is considered not to have been used.

[0083] (2) Set generation

[0084] Used Solution Set: This set includes solutions that were actually used in the problem-solving information and successfully matched a standard solution in the solution set. Each used solution is associated with its usage frequency, matching degree, and corresponding problem category information.

[0085] Unused Solution Set: This set includes solutions from the solution set that are not matched with the problem information or whose matching degree is below a preset threshold. This set contains feasible solutions that students have not yet used or rarely use, providing potential material for subsequent problem generation.

[0086] (3) Result verification

[0087] Cross-validation is performed on the generated sets of used and unused solutions to ensure the accuracy and completeness of the classification and avoid misclassification of solutions due to feature noise or matching errors.

[0088] Output the following results for subsequent processing: Used solution set: contains a list of solutions actually used by students in the historical problem-solving process. Each solution is accompanied by its corresponding question category, usage identifier, and matching feature summary; Unused solution set: contains a list of solutions in the solution set that have not been used by students. Each solution is also associated with question category and feature information, providing a data foundation for subsequent screening and question generation.

[0089] Regarding step S4:

[0090] Obtain the data from step S3: Used solution set: contains a list of solutions actually used by students in the historical problem-solving process, with each solution associated with its corresponding question category and usage record; Unused solution set: contains a list of solutions that have not been used by students.

[0091] And preset usage threshold: a numerical parameter predefined by the system to determine whether the solution is being used sufficiently. This threshold is set based on historical teaching data or expert experience.

[0092] The statistical analysis and update operations for the set of used solutions are performed as follows:

[0093] (1) Usage value statistics

[0094] For each solution in the set of used solutions, count the number of times it appears in the set of student historical accounting document solutions. This count is used as the usage value of the solution. The usage value is a numerical indicator that reflects the frequency with which the solution is used in the student's problem-solving process.

[0095] For example, for the "direct method of debit and credit balance" solution, the total number of times all students used this solution in "voucher filling" type questions is counted as its usage value.

[0096] (2) Threshold comparison and set update

[0097] The usage value of each solution is compared with a preset usage threshold: if the usage value is lower than the preset usage threshold, the solution is determined to be underutilized, removed from the set of used solutions, and added to the set of unused solutions; if the usage value is greater than or equal to the preset usage threshold, the solution is retained in the set of used solutions.

[0098] For example, if the preset usage threshold is 5 times, and a certain solution only appears 3 times in the historical solution set, then it will be moved to the unused solution set.

[0099] (3) Collection integrity maintenance

[0100] Ensure that during the set update process, the metadata such as problem category association information and feature data of all solutions remain intact and are transferred along with the solutions.

[0101] The updated set of used solutions and the set of unused solutions are deduplicated and sorted to ensure data consistency.

[0102] Output the following results for subsequent processing: Updated set of used solutions: contains a list of solutions whose usage values ​​have reached or exceeded the preset usage threshold, with each solution accompanied by its usage value statistics; Updated set of unused solutions: contains the original set of unused solutions and low-usage solutions transferred from the set of used solutions, together forming the optimized pool of unused solutions.

[0103] Regarding step S5:

[0104] Obtain the data from step S4: Updated set of unused solutions: contains a list of solutions that are not fully used or not used, and each solution is associated with its question category and characteristic information; Historical accounting document solution set: a complete set of students' historical solution records stored in the system, containing solution information, solutions used and associated data for all question categories.

[0105] The effectiveness evaluation and ranking optimization of the unused solution set are performed as follows:

[0106] (1) Frequency statistics

[0107] Iterate through each solution in the set of unused solutions, perform a global search and match in the set of historical accounting document solutions, and count the total number of times each solution appears in the historical solution records of all students.

[0108] The frequency of occurrence indicates the adoption of the solution in the historical teaching environment, serving as an objective indicator for evaluating the effectiveness of the solution.

[0109] (2) Efficiency calculation and assignment

[0110] The frequency of occurrence obtained from statistics is directly defined as the effectiveness rate of the solution. Effectiveness rate, as a numerical evaluation indicator, reflects the applicability and popularity of the solution in history teaching practice.

[0111] For example, if a certain solution appears 20 times in the historical solution set, its effectiveness rate is assigned a value of 20.

[0112] (3) Descending order sorting

[0113] Based on the efficiency values ​​of each solution, all solutions in the unused solution set are sorted in descending order.

[0114] After arrangement, an ordered sequence of solutions is formed, with highly efficient solutions at the beginning of the sequence and less efficient solutions at the end.

[0115] This sorting process ensures that unused solutions with high teaching value and high historical usage frequency are given priority.

[0116] (4) Data preprocessing

[0117] Duplicate records are removed during the statistical process to ensure the accuracy of the frequency count.

[0118] For solutions with the same efficiency, secondary sorting is performed according to solution number or problem category to ensure the deterministic nature of the sequence structure.

[0119] Output the following results for subsequent processing: Unused solution set sorted in descending order: contains a sequence of unused solutions sorted from highest to lowest efficiency, with each solution accompanied by its efficiency value and complete attribute information.

[0120] Regarding step S6:

[0121] For example, students' learning levels are divided based on the average score of their historical accounting document problem-solving set within a specified time window, and a predetermined number of unused solutions from the solution set are retained in descending order based on the learning level, including:

[0122] Extract all problem-solving records within the specified time window from the student's historical accounting document problem-solving collection;

[0123] Calculate the average score of all student problem-solving records within the specified time window;

[0124] The calculated average score is matched with the preset learning level division method to determine the student's learning level;

[0125] Based on the determined learning level, obtain the number of unused solutions to be retained for that level from the preset parameters;

[0126] Based on a predetermined number of solutions to retain, starting from the beginning of the unused solution set after descending order, a corresponding number of solutions are retained sequentially.

[0127] Furthermore, such as Figure 2 As shown, the validity verification of the unused solution set includes:

[0128] For unused solutions within the same question category, cluster analysis is performed based on their frequency of occurrence in different students' historical accounting document solution sets;

[0129] If the usage rate of a certain type of solution in the set of solutions to a set of historical accounting documents exceeds a set threshold, then the solution of that type will be included in the reserved candidate set corresponding to each learning level.

[0130] If the difference in usage rate between different learning levels and two different question categories exceeds a preset difference threshold, then the solution will be removed from the retention candidate set.

[0131] Determine if the number of retained solutions for each learning level is zero. If so, include the most efficient solution within the current calibration time window into the retained candidate set for that learning level, and use the current retained candidate set as the set of unused solutions.

[0132] Furthermore, the set of verified unused solutions is dynamically updated, including:

[0133] The system retrieves newly generated problem-solving records from students and updates the historical accounting document problem-solving set. Based on the updated historical accounting document problem-solving set, it recalculates the usage value of each solution and includes solutions with usage values ​​below a preset usage threshold in the unused solution set.

[0134] Through the above technical solutions, this application constructs a closed-loop optimization mechanism that does not utilize a set of solutions. Based on cluster analysis and multi-dimensional threshold determination, it achieves intelligent screening and verification of potentially effective solutions; through cross-category applicability testing, it ensures the stability and reliability of solution recommendations; and with the help of a dynamic update strategy, the system can continuously optimize solution configuration based on the latest learning data. This mechanism not only guarantees the comprehensiveness and accuracy of solution discovery but also maintains the timeliness of the solution set through real-time updates, improving the accuracy of personalized question generation. Furthermore, the closed-loop optimization architecture enables the system to have self-improvement capabilities, effectively enhancing the adaptability and sustainable development capabilities of the accounting document question generation system.

[0135] By assessing students' abilities and using personalized selection, a subset of unused solutions is generated for different learning levels.

[0136] Specifically, the acquired data includes: Student historical accounting document problem-solving set: records all accounting document test questions solved by the target student within a historical time period, including question scores, solution time, and question category information; Defined time window: a system-preset specific time period parameter (e.g., the most recent 30 days or this semester) used to limit the time range for learning ability assessment; Unused solution set sorted in descending order: an ordered set derived from the S5 step output, containing unused solutions sorted in descending order of effectiveness, with each solution accompanied by an effectiveness value and question category attribute; Learning level classification rules: a system-predefined student ability grading standard, including the correspondence between score ranges and levels, and the number of solutions to be retained for each learning level: a system-preset parameter for the number of unused solutions to be retained for different learning levels.

[0137] The solution retention process is completed through student ability assessment and personalized selection, as follows:

[0138] (1) Learning ability assessment

[0139] Extract all problem-solving records within the specified time window from the students' historical accounting document problem-solving collection.

[0140] Calculate the average score of all student solutions within the time window using the formula: Total score / Number of problems solved.

[0141] The calculated average score is matched with the preset learning level classification rules to determine the student's learning level.

[0142] For example, the preset rules are: an average score of ≥90 is "Grade A", 80-89 is "Grade B", 70-79 is "Grade C", and <70 is "Grade D".

[0143] (2) Determining the number of solutions to retain

[0144] Based on the determined learning level, the number of unused solutions to be retained for that level is obtained from the system's preset parameters.

[0145] For example, the system presets: 8 solutions are retained for level A, 6 solutions for level B, 4 solutions for level C, and 2 solutions for level D.

[0146] (3) Personalized solution selection

[0147] Based on a predetermined number of solutions to retain, starting from the beginning of the unused solution set after descending order, a corresponding number of solutions are retained sequentially.

[0148] The retention process ensures that efficient solutions are prioritized, while also considering the balanced distribution of solutions across different problem categories.

[0149] For example, for a student who is rated B, the six most efficient solutions are retained from the set of solutions they did not use.

[0150] (4) Data validation

[0151] Verify whether the retained subset of solutions meets the predetermined quantity requirement.

[0152] Check whether the retained solutions cover the main problem categories and ensure the representativeness of the knowledge points.

[0153] Output the following results for subsequent processing: Set of unused solutions: contains a predetermined number of unused solutions selected based on the student's learning level. These solutions form the basis for generating targeted questions later; Student learning level information: records the student's level assessment results and corresponding average score data for subsequent analysis and system optimization.

[0154] Regarding step S7:

[0155] The acquired data includes: an accounting document question bank: the system's basic question resources, containing multiple accounting document questions, each associated with a question category, question stem information, knowledge point, and difficulty level; a set of unused solutions after retention: derived from the filtering results of step S6, containing a predetermined number of unused solutions retained according to the student's learning level, each solution associated with its question category and feature information; a question category system: the system's predefined accounting document question classification standards, such as "voucher preparation," "ledger registration," and "report preparation," etc.; and a knowledge point and difficulty level system: the system's established knowledge point classification framework and difficulty grading standards, used for the feature description of questions.

[0156] The generation of targeted questions and template construction are achieved through multiple steps, as follows:

[0157] (1) Targeted question generation

[0158] Iterate through each solution in the set of unused solutions that are retained, and match and combine them with the basic questions in the accounting document question bank.

[0159] For each target solution, select a basic test question that matches its question category as a prototype.

[0160] By reconstructing the question information, resetting parameters, or changing conditions, the target solution is integrated into basic test questions to generate new targeted questions.

[0161] For example, for the unused solution of "indirect method of multi-column ledger registration", select the corresponding basic test questions of ledger registration, adjust its business scenario and registration requirements, and generate targeted questions that reflect the characteristics of this solution.

[0162] (2) Grouping by question type

[0163] All generated targeted questions are grouped according to their respective question categories to form multiple sets of document test questions.

[0164] Each group of document test questions contains multiple targeted questions under the same question category, such as all voucher filling questions forming one group.

[0165] During the grouping process, ensure that the number of questions in each group is reasonably distributed and covers the core teaching content of that category.

[0166] (3) Construction of test question templates

[0167] For each group's set of test questions, analyze the related knowledge points and difficulty levels:

[0168] Key knowledge points: Identify the core accounting concepts, skill requirements, or business scenarios involved in the questions.

[0169] Difficulty level: The difficulty of a problem is assessed based on factors such as problem complexity, number of calculation steps, and the degree of concealment of conditions.

[0170] Based on the logical connections and progressive difficulty of the knowledge points, the group-specific test questions are grouped and arranged to form a group-specific test question template.

[0171] Each group's test template includes: a sequence of knowledge points for that question category; a distribution of difficulty levels; an ordered arrangement of targeted questions; and a correspondence between questions and solutions.

[0172] (4) Template optimization and verification

[0173] Check the comprehensiveness of the knowledge points covered in the test question templates for each group to ensure that no important teaching content is omitted.

[0174] Verify the gradual increase in difficulty levels to ensure that the difficulty distribution of questions within the template conforms to teaching principles.

[0175] Confirm the accuracy of the correspondence between targeted questions and solutions not used to ensure clear training objectives.

[0176] Output the following results for subsequent processing: Group Question Template Set: Contains multiple question templates organized by question category. Each template contains a complete knowledge point structure, difficulty level distribution, and targeted question sequence; Targeted Question Bank: Records all generated targeted questions and their attribute information as alternative resources for test paper compilation.

[0177] Regarding step S8:

[0178] For example, dynamically configuring group-specific test question templates and using machine learning optimization algorithms for test paper generation includes:

[0179] Based on the test paper structure requirements, a set of candidate questions that meet the conditions is extracted from the group test question template set;

[0180] A machine learning optimization method based on genetic algorithms is adopted to intelligently filter and combine questions from the candidate question set, and the final candidate question set is obtained by iteratively calculating and optimizing the question combination.

[0181] The final set of candidate questions will be used as the accounting document exam paper.

[0182] Specifically, the acquired data includes: a set of group-specific question templates: structured question resources derived from the output of step S7, containing multiple question templates organized by question category. Each template includes a sequence of knowledge points, a distribution of difficulty levels, and a targeted arrangement of questions; test paper structure requirements: system-preset parameters for test paper composition, including the total number of questions, the distribution ratio of question categories, the scope of knowledge point coverage, the range of difficulty coefficients, and the constraints on test duration; and test paper optimization objectives: system-defined optimization indicators, including maximizing knowledge point coverage, ensuring reasonable difficulty distribution, and providing diverse questions and comprehensive solution coverage.

[0183] Intelligent test paper generation is achieved through dynamic configuration and machine learning optimization. The specific process is as follows:

[0184] (1) Dynamic configuration initialization

[0185] Based on the test paper structure requirements, a set of candidate questions that meet the conditions is extracted from the group test question template set.

[0186] Based on the distribution ratio of question categories, determine the allocation of the number of questions in each category in the test paper.

[0187] Initialize the test paper structure framework, and set the number of questions, scores, and order of each section.

[0188] (2) Machine learning optimization test paper generation

[0189] We employ machine learning optimization algorithms based on genetic algorithms, particle swarm optimization, or constraint satisfaction to intelligently filter and combine candidate problems from the pool.

[0190] The optimization process is guided by the goal of test paper optimization and performs multi-objective optimization through the following dimensions: Knowledge point coverage optimization: ensuring that the test paper covers the core knowledge points of accounting documents and avoiding the omission of important content; Difficulty balance optimization: controlling the overall difficulty coefficient of the test paper within the preset range and maintaining a reasonable difficulty gradient; Question diversity optimization: ensuring a balanced presentation of different solutions and avoiding repetition or bias in solutions; Question type structure optimization: maintaining an appropriate ratio of objective questions to subjective questions to meet the requirements of teaching assessment; The algorithm continuously optimizes the question combination through iterative calculation, evaluates the fitness score of each combination, and gradually approaches the optimal solution.

[0191] (3) Verification of the integrity of the test paper

[0192] Perform a completeness check on the optimized test paper to verify whether it meets all preset constraints.

[0193] The inspection includes: the total number of questions meets the standard, the distribution of question types is consistent, the coverage of knowledge points is complete, and the distribution of difficulty is reasonable.

[0194] Make partial adjustments and further optimizations to test papers that do not meet the requirements to ensure output quality.

[0195] (4) Examination paper formatting:

[0196] The finalized questions will be formatted and organized according to the standard exam paper format.

[0197] Add necessary exam instructions, score markings, and answer guidelines.

[0198] Generate a unified test paper number and timestamp information.

[0199] Output the following results: Accounting Document Exam Paper: A formal exam paper document containing complete exam questions, standard answers, scoring criteria, and exam instructions; Exam Paper Metadata: Data that records the composition and analysis of the exam paper, including auxiliary information such as knowledge point coverage statistics, difficulty distribution analysis, and solution usage.

[0200] This invention also verifies the validity of accounting document test papers, including:

[0201] Obtain the student's answer results and learning level for the corresponding accounting document exam;

[0202] The test results are compared and analyzed with the learning level to calculate the test paper matching index. If the test paper matching index is lower than the preset matching threshold, the learning level division method or the number of solutions retained is adjusted based on the test results.

[0203] Through the aforementioned technical solution, this application achieves the validity verification of accounting document exam papers. By comparing and analyzing students' answers with their learning levels in real time, calculating the exam paper matching index, and dynamically adjusting the learning level division method or the number of solutions retained, it ensures that the generated exam papers are highly matched with students' actual learning abilities. Through intelligent judgment and feedback adjustment of the matching threshold, it not only ensures the adaptability of the exam paper difficulty but also improves the accuracy of personalized training, effectively avoiding the problem of mismatch between questions and student levels in traditional exam paper generation methods. Furthermore, the dynamic adjustment mechanism enhances the system's self-optimization capability, enabling the exam paper generation process to continuously improve as students progress in their learning, thereby improving the overall effectiveness and efficiency of accounting document teaching.

[0204] like Figure 3 The diagram shows the data flow of the dynamically configured question generation and test paper assembly method for accounting documents.

[0205] To facilitate understanding of the above embodiments, a specific application scenario of the above embodiments will be used as an example for illustration below:

[0206] Scenario: Wang, an accounting student, is using the "Smart Accounting Training Platform" for specialized training in accounting document processing. The platform has generated grouped test templates based on Wang's historical problem-solving data (such as voucher preparation and ledger entry). These templates are built upon solutions she has not fully utilized (such as the "red ink reversal method" and "multi-column ledger entry"), and are grouped by question type (such as voucher processing and ledger entry), including knowledge point sequences, difficulty level distribution, and targeted questions.

[0207] (1) Dynamic configuration initialization

[0208] The platform extracts a candidate set of questions from the group question template set according to the test paper structure requirements:

[0209] Exam paper structure requirements:

[0210] Total number of questions: 20;

[0211] Question categories: Voucher preparation (40%), ledger entry (40%), report preparation (20%).

[0212] Knowledge points covered: Must cover core knowledge points such as "debit and credit balance principle", "multi-column ledger entry", and "red ink reversal method".

[0213] Difficulty level range: 0.6-0.8 (medium difficulty);

[0214] Exam duration: 60 minutes;

[0215] Initialization process:

[0216] The platform extracts a candidate set of questions that meet the above conditions from the group-specific question template set, including:

[0217] Voucher filling template: Includes 8 questions, covering knowledge points such as the "debit and credit balance principle" and the "red ink reversal method".

[0218] Ledger Registration Template: Contains 8 questions covering knowledge points such as "multi-column ledger registration" and "account classification".

[0219] Report preparation template: Includes 4 questions covering knowledge points such as "profit and loss statement preparation" and "balance calculation".

[0220] Initialize the exam paper framework: set the number of questions for each section (8 questions on voucher filling, 8 questions on ledger registration, and 4 questions on report preparation), the score (5 points per question, 100 points in total), and the order (voucher filling first, followed by ledger registration, and finally report preparation).

[0221] (2) Machine learning optimization test paper generation

[0222] The platform employs a machine learning optimization method based on genetic algorithms for intelligent test paper generation.

[0223] Optimization goal:

[0224] Maximize knowledge coverage: Ensure the exam covers all core knowledge points.

[0225] Reasonableness of difficulty distribution: The overall difficulty coefficient is controlled between 0.6 and 0.8.

[0226] Problem diversity: Avoid repetitive solutions and ensure a balanced presentation of solutions such as the "red ink cancellation method" and "multi-column ledger entry".

[0227] Question structure optimization: Objective questions (multiple choice) account for 60%, and subjective questions (calculation questions) account for 40%.

[0228] Optimization process:

[0229] Genetic algorithm initialization: Randomly generate multiple combinations of questions (chromosomes), each combination containing 20 questions.

[0230] Iterative calculation:

[0231] The fitness score for each combination is evaluated based on knowledge coverage (weight 40%), difficulty matching (weight 30%), solution diversity (weight 20%), and question type ratio (weight 10%).

[0232] Selection, crossover, and mutation: retain combinations with high fitness and generate new combinations through crossover and mutation operations.

[0233] After 50 iterations, the algorithm converged to an optimal combination of problems, achieving a fitness score of 92 out of 100.

[0234] Output the final candidate set of questions: containing 20 questions, of which:

[0235] There are 8 questions on voucher preparation, including 4 questions involving the "red ink cancellation method".

[0236] There are 8 questions on ledger entry: including 4 questions on "multi-column ledger entry".

[0237] Four questions on report preparation: covering the knowledge points of "profit and loss statement preparation".

[0238] (3) Verification of the integrity of the test paper

[0239] The platform verifies the optimized test papers:

[0240] Inspection content:

[0241] Total number of questions: 20, which meets the requirements.

[0242] Question categories distribution: 8 questions on voucher preparation (40%), 8 questions on ledger entry (40%), and 4 questions on report preparation (20%), which meets the requirements.

[0243] Knowledge coverage: All core knowledge points are covered without omission.

[0244] Difficulty distribution: The average difficulty coefficient is 0.72, which is within the preset range.

[0245] Partial adjustment: A voucher filling question was found to have a difficulty coefficient of 0.85, which is slightly higher than the upper limit. The platform automatically replaced it with a backup question with a difficulty of 0.7.

[0246] Exam paper format:

[0247] Layout and organization: Group by question type, and add exam instructions, score markings, and answer guidelines.

[0248] Generate metadata: The exam paper number is “ACC20231001”, the timestamp is “20231001”, and records the knowledge point coverage statistics (coverage rate 95%), difficulty analysis (average difficulty 0.72) and solution usage (“red ink cancellation method” used 4 times, “multi-column ledger registration” used 4 times).

[0249] (4) Verification of the validity of accounting documents test papers

[0250] The platform verifies the validity of the generated test paper based on Wang's answers:

[0251] Data Acquisition:

[0252] After Wang completed the test, the platform collected her answer results: 75 points (15 correct answers and 5 incorrect answers).

[0253] Wang's current learning level is "Advanced Stage" (based on his historical average score of 85).

[0254] Comparative analysis:

[0255] The test paper matching index is calculated based on the comparison between the test results and the expected performance at the learning level. The formula is:

[0256]

[0257] The expected score was based on the learning level (the expected score for the "advanced stage" was 80-90 points). Wang's actual score was 75 points, which was lower than the lower limit of the expected score, and the matching degree was 83.3% (75 / 90).

[0258] Preset matching threshold: 85%. Since 83.3% < 85%, the matching degree is lower than the threshold.

[0259] Dynamic adjustment:

[0260] The platform adjusted the learning level classification method based on the answer results: Wang made many mistakes in the voucher filling questions, indicating that she did not have a good grasp of the "red ink cancellation method". Therefore, the platform temporarily downgraded her learning level from "advanced stage" to "consolidation stage".

[0261] Adjust the number of solutions to retain: For the "consolidation phase", the number of solutions to retain is reduced from 6 to 4, with priority given to retaining basic solutions (such as the "direct method of borrowing and lending balance") and reducing the retention of advanced solutions (such as the "red-ink cancellation method").

[0262] Update the unused solution set: The platform recalculates the usage value of Wang's solution, marks the usage value of "red cancellation method" as 0 (unused), and includes it in the unused solution set for the next test paper assembly.

[0263] Example 2

[0264] This invention also provides a dynamically configurable accounting document question generation and test paper assembly system for implementing the above-described dynamically configurable accounting document question generation and test paper assembly method, comprising:

[0265] The data acquisition module 100 is used to acquire the sequence of solution information from accounting document test questions and the student's historical accounting document solution set. The solution information includes question category, question stem information, solution information, and a set of solutions related to question category and question stem information.

[0266] The feature extraction module 200 is used to extract features from the problem-solving information and the solutions in the solution set using machine learning, generating common features and distinctive features. The common features are those shared by both, and the distinctive features are the unique features of the problem-solving information.

[0267] The feature comparison module 300 is used to compare the common features and distinguishing features of the solutions extracted from the problem-solving information with the corresponding solution set, and generate the set of used solutions and the set of unused solutions;

[0268] The solution set update module 400 is used to count the usage value of each solution in the used solution set. If the usage value is lower than the preset usage threshold, the solution is included in the unused solution set.

[0269] The solution set descending order module 500 is used to count the number of times each solution in the unused solution set appears in the historical accounting document solution set, and uses the number of occurrences as the effectiveness rate to sort the unused solution set in descending order according to the effectiveness rate;

[0270] The solution set retention module 600 is used to classify students’ learning levels based on the average score of the students’ historical accounting document solution set within a specified time window, and retain a predetermined number of unused solutions in the solution set in descending order based on the learning level.

[0271] The test question template acquisition module 700 is used to generate targeted questions from accounting document test questions based on the retained set of unused solutions, group the targeted questions by category to obtain group document test questions, and combine them based on the knowledge points and difficulty levels of the group document test questions to obtain group test question templates.

[0272] The test paper generation module 800 is used to dynamically configure the test question templates for different groups and use machine learning optimization algorithms to generate test papers for accounting documents.

[0273] Example 3

[0274] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method for dynamically configuring accounting document question generation and test paper preparation.

[0275] Example 4

[0276] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is configured to perform a dynamically configured method for generating and compiling accounting document questions.

[0277] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be constructed using software plus necessary general-purpose hardware platforms, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or certain parts of embodiments.

[0278] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A method for dynamically configuring an accounting document to set a test paper, characterized in that, The method comprises the following steps: obtaining accounting document test questions and a sequence of problem solving information in a student historical accounting document problem solving set, wherein the problem solving information comprises a question category, a question stem information, problem solving information, and a solution set associated with the question category and the question stem information; using machine learning to extract features from the problem solving information and the solution set, to generate common features and distinguishing features, wherein the common features are common features of both, and the distinguishing features are unique features of the problem solving information; comparing the solution of the problem solving information and the solution set to generate a used solution set and an unused solution set; counting the use value of each solution in the used solution set, and if the use value is lower than a preset use threshold, the solution is included in the unused solution set; counting the number of occurrences of each solution in the unused solution set in the historical accounting document problem solving set, and taking the number of occurrences as the efficiency, and arranging the unused solution set in descending order according to the efficiency; dividing the learning level of the student based on the average score of the student historical accounting document problem solving set within a calibration time window, and reserving a predetermined number of solutions in the unused solution set after descending arrangement based on the learning level; generating targeted questions based on the accounting document test questions and the reserved unused solution set, grouping the targeted questions by category to obtain group document test questions, and combining the group document test questions based on the knowledge points and difficulty levels to obtain group test templates; dynamically configuring the group test templates and using a machine learning optimization algorithm to process the test paper to obtain an accounting document test paper.

2. The method for dynamically configuring the accounting document setting question group according to claim 1, wherein, The process of using machine learning to extract features from the problem solving information and the solution set to generate common features and distinguishing features comprises: obtaining the text description of the question stem information and the document format template; extracting the keywords, semantic vectors and syntax structures of the question stem information through natural language processing technology; encoding each solution in the solution set to extract its method type, applicable conditions, calculation logic and step sequence features; calculating the similarity and pattern matching of the solution features of the problem solving information and the solution features of the solution set; taking the common features existing in the problem solving information and the solution set as common features, and taking the features unique to the problem solving information and not included in the solution set as distinguishing features.

3. The method for dynamically configuring the accounting document setting question group according to claim 1, wherein, The process of dividing the learning level of the student based on the average score of the student historical accounting document problem solving set within a calibration time window, and reserving a predetermined number of solutions in the unused solution set after descending arrangement based on the learning level comprises: extracting all problem solving records within the calibration time window from the student historical accounting document problem solving set; calculating the average score of all problem solving records within the calibration time window; matching the calculated average score with a preset learning level division method to determine the learning level of the student; according to the determined learning level, obtaining the number of unused solutions corresponding to the level from the preset parameters; based on the determined number of reservations, starting from the starting position of the descending arranged unused solution set, sequentially reserving the corresponding number of solutions.

4. The method of claim 3, wherein, The method further comprises verifying the effectiveness of the unused solution set, which comprises: The unused solution under the same topic category is clustered according to the frequency of occurrence in different student historical accounting document problem solving sets; If the usage rate of a certain solution in a preset number of historical accounting document problem solving sets exceeds a preset threshold, the solution is included in the reserved candidate set corresponding to each learning level; If the usage rate difference of the solutions reserved by different learning levels in two different topic categories exceeds a preset difference threshold, the solution is excluded from the reserved candidate set; If the number of reserved solutions corresponding to each learning level is zero, the solution with the highest efficiency in the current calibration time window is included in the reserved candidate set of the learning level, and the reserved candidate set at this time is taken as the unused solution set.

5. The method of claim 4, wherein, The unused solution set after verification is dynamically updated, including: Obtain the newly generated problem solving records of students and update the historical accounting document problem solving set, and based on the updated historical accounting document problem solving set, the usage value of each solution is recalculated, and the solution with a usage value lower than a preset usage threshold is included in and updated in the unused solution set.

6. The method for dynamically configuring the accounting document setting question group according to claim 1, wherein, The dynamic configuration of the group test question template and the use of the machine learning optimization algorithm for group test processing specifically include: According to the requirements of the test paper structure, the question candidate set meeting the conditions is extracted from the group test question template set; A machine learning optimization method based on genetic algorithm is used to screen and combine from the question candidate set, and the final question candidate set is obtained through iterative calculation and optimization of the question combination. The final question candidate set is taken as the accounting document test paper.

7. The method of claim 6, wherein the accounting document is configured dynamically. The effectiveness of the accounting document test paper is verified, including: Obtain the student's answer to the corresponding accounting document test paper and the learning level; Compare and analyze the answer and the learning level, calculate the test paper matching degree index, and if the test paper matching degree index is lower than the preset matching threshold, adjust the learning level division method or the number of solution reservations based on the answer.

8. A dynamically configured accounting document setting and grouping system, characterized by, A dynamic configuration of an accounting document question setting and test paper group method is implemented, including: A data acquisition module is used to acquire accounting document test questions and problem solving information sequences in student historical accounting document problem solving sets, the problem solving information including topic categories, question stem information, problem solving information, and solution set associated with topic categories and question stem information; A feature extraction module is used to extract features of problem solving information and solutions in the solution set using machine learning to generate common features and distinguishing features, the common features being common features of both, and the distinguishing features being unique features of the problem solving information; A feature comparison module is used to compare the common features and distinguishing features of the solution extracted from the problem solving information and the corresponding solution set to generate a used solution set and an unused solution set; A solution set update module is used to calculate the usage value of each solution in the used solution set, and if the usage value is lower than a preset usage threshold, the solution is included in the unused solution set; A solution set descending order module is used to count the number of occurrences of each solution in the unused solution set in the historical accounting document problem solving set, and the number of occurrences is taken as the efficiency, and the unused solution set is arranged in descending order according to the efficiency; The solution set reservation module is configured to divide the learning level of the student based on the average score of the solution set of the historical accounting document of the student within the calibration time window, and reserve a predetermined number of solutions in the unused solution set in descending order based on the learning level; The test question template acquisition module is configured to generate targeted questions based on the unused solution set after reservation, group the targeted questions by category to obtain group document test questions, and combine the group document test questions based on the knowledge points and difficulty levels to obtain group test question templates; The test paper processing module is configured to dynamically configure the group test question templates and use a machine learning optimization algorithm to process the test paper to obtain an accounting document test paper. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the dynamic configuration accounting document question setting and test paper processing method of any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to implement the dynamic configuration accounting document question setting and test paper processing method of any one of claims 1-7.

Images