Proposition method and system combining large model and inference engine, medium and product

Abstract

The invention discloses a proposition method and system combining a large model and an inference engine, a medium and a product, and relates to the field of artificial intelligence. Matching knowledge point rules from a preset reasoning rule base; combining the knowledge point rules to obtain a knowledge point combination sample; reasoning the knowledge point combination topics through a reasoning engine to generate a formal language expression topic set; based on a preset translation template, converting the formal language expression question set into a natural language expression question set; rewriting the natural language expression question set through the large model to generate an optimized question set; judging whether each optimization question in the optimization question set meets a preset output condition or not; and outputting the optimization questions meeting a preset output condition in the optimization question set. By implementing the method, the logic accuracy of question generation is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to proposition methods, systems, media and products combining large models and reasoning engines. Background Art

[0002] With the rapid development of artificial intelligence, the application of automatic question generation technology in the field of education is becoming more and more extensive. How to use automatic question generation technology to quickly generate questions that meet the requirements of the syllabus while ensuring the quality and diversity of the questions has become an important challenge faced by educational technology developers.

[0003] At present, a large number of known math problems are stored in the automated question generation system. The automated question generation system can generate new problems by dynamically replacing numbers or problem situations in known math problems based on problem generation requirements.

[0004] Although relevant technologies can quickly generate a large number of questions in some cases, due to the strict requirements of mathematics on calculation and reasoning, it is often difficult to ensure the logical correctness of the generated questions by simply replacing and rewriting known mathematical questions. Summary of the invention

[0005] The present application provides a proposition method, system, medium and product combining a large model and a reasoning engine, which are used to improve the logical accuracy of question generation.

[0006] In a first aspect, the present application provides a proposition method combining a large model and an inference engine, which is applied to a proposition system, the method comprising: receiving proposition input information, the proposition input information including proposition requirements; matching knowledge point rules whose relevance to the proposition requirements exceeds a preset relevance threshold from a preset inference rule library; combining the knowledge point rules to obtain knowledge point combination sample questions; reasoning the knowledge point combination sample questions through an inference engine to generate a formal language expression question set, the formal language expression question set including multiple formal language expression questions; based on a preset translation template, converting the formal language expression question set into a natural language expression question set, the preset translation template including a correspondence between formal language expressions and natural language expressions, the natural language expression question set including multiple natural language expression questions; rewriting the natural language expression question set through a large model to generate an optimized question set, the optimized question set including multiple optimized questions; judging whether each optimized question in the optimized question set meets a preset output condition, the preset output requirement including question grammatical accuracy, proposition requirement consistency, and question answer validity; and outputting the optimized questions in the optimized question set that meet the preset output condition.

[0007] By adopting the above technical solutions, the proposition system has established a complete proposition process through the reasoning engine and the big model. This combination ensures that the automatically generated questions not only have logical accuracy, but also have natural and fluent language expression. In addition, through multiple verification mechanisms (question grammatical accuracy, consistency of proposition requirements, and validity of question answers), the quality of the final question output is ensured, overcoming the problem that simply replacing numbers or problem situations in known math questions to generate questions may lead to logical errors, thus improving the quality and usability of automatic proposition.

[0008] In combination with some embodiments of the first aspect, in some embodiments, the knowledge point combination sample questions are inferred by an inference engine to generate a set of questions expressed in a formal language, which specifically includes: extracting key elements and logical relationships in the knowledge point combination sample questions; based on the inference engine, formalizing the key elements and the logical relationships to construct a formal language grammar structure tree; traversing different branch paths of the formal language grammar structure tree to generate multiple formal language expression variants; and aggregating the formal language expression variants to obtain the set of questions expressed in the formal language.

[0009] By adopting the above technical solution, the question setting system extracts the key elements and logical relationships in the knowledge point combination sample questions, constructs a formal language syntax structure tree, and traverses different branch paths to generate multiple formal language expression variants to obtain a formal language expression question set. This formal processing method ensures the rigor and integrity of the questions at the logical level, increases the diversity of the questions, and greatly improves the controllability and logical accuracy of question generation.

[0010] In combination with some embodiments of the first aspect, in some embodiments, the natural language expression topic set is rewritten by a large model to generate an optimized topic set, specifically including: splicing preset rewriting prompt words with each natural language expression topic in the natural language expression topic set to obtain multiple splicing results; inputting the splicing results into the large model to obtain an optimized topic after the natural language expression topic is rewritten; and adding the optimized topic to the optimized topic set.

[0011] By adopting the above technical solution, the question setting system will splice the preset rewriting prompt words with the natural language expression questions, and then input them into the big model, giving full play to the advantages of the big model in natural language processing. By rewriting each natural language expression question, you can get a more natural and smooth optimized question that meets the needs of educational practice. This rewriting process not only maintains the core meaning and logical relationship of the question, but also through the language optimization capabilities of the big model, it can make the question expression closer to actual teaching needs, improving the readability and educational effect of the question.

[0012] In combination with some embodiments of the first aspect, in some embodiments, the determination of whether each optimized question in the set of optimized questions meets preset output conditions, the preset output requirements including question grammatical accuracy, question requirement consistency and question answer validity, specifically includes: performing syntactic analysis on the optimized question to obtain a syntactic analysis result; based on the syntactic analysis result, determining whether the optimized question meets the preset question grammatical accuracy standard; calculating the similarity between the knowledge point feature vector of the optimized question and the knowledge point feature vector of the question requirement; determining whether the similarity exceeds a preset similarity threshold to determine whether the optimized question meets the preset question requirement consistency standard; solving the optimized question to obtain a solution result; based on the solution result, determining whether the optimized question meets the preset question answer validity standard.

[0013] By adopting the above technical solutions, the question setting system controls the three dimensions of question grammatical accuracy, question setting requirements consistency, and question answer validity, and establishes a comprehensive question quality verification mechanism. This multi-dimensional question verification method can effectively filter out unqualified questions, ensure the quality of output questions, make the question screening process more scientific and reliable, and significantly improve the practicality of the question setting system.

[0014] In combination with some embodiments of the first aspect, in some embodiments, after the step of determining whether each optimized question in the optimized question set meets preset output conditions, wherein the preset output requirements include question grammatical accuracy, consistency of proposition requirements, and validity of question answers, the method further includes: performing error type analysis on the optimized questions that do not meet the preset output conditions; selecting corresponding correction prompt words from a preset correction rule library based on the error type; and inputting the correction prompt words into the large model to obtain a corrected optimized question.

[0015] By adopting the above technical solutions, the question-setting system introduces an error type analysis and correction mechanism, achieving closed-loop optimization of question quality. For questions that do not meet the preset output conditions, the question-setting system will analyze the specific error type, select the corresponding correction prompt words from the preset correction rule library, and make corrections through the large model. This precise correction method based on error analysis not only improves the success rate of question generation, but also continuously optimizes the performance of the question-setting system by accumulating error types and correction experience.

[0016] In combination with some embodiments of the first aspect, in some embodiments, outputting the optimized questions in the optimized question set that meet the preset output conditions specifically includes: extracting difficulty characteristic parameters of the optimized questions in the optimized question set that meet the preset output conditions, the difficulty characteristic parameters including the number of knowledge points, the difficulty coefficient of the knowledge points, and the number of problem-solving steps; determining the difficulty score of the optimized question based on the number of knowledge points, the difficulty coefficient of the knowledge points, and the number of problem-solving steps; and outputting the optimized questions in the optimized question set that meet the preset output conditions in order of the difficulty scores.

[0017] By adopting the above technical solution, the question setting system introduces a question sorting mechanism based on difficulty characteristic parameters, so that the question difficulty can be scientifically evaluated and the questions can be output according to the difficulty level. This question difficulty evaluation method based on multi-dimensional parameters makes the generated questions more suitable for the gradient training needs in educational practice, and improves the pertinence of question application and teaching effect.

[0018] In combination with some embodiments of the first aspect, in some embodiments, after the step of receiving proposition input information, wherein the proposition input information includes proposition requirements, the method further includes: retrieving candidate questions in a preset question bank whose matching degree with the proposition requirements exceeds a preset matching degree threshold; determining whether the candidate question meets the preset output conditions, wherein the preset output requirements include question grammatical accuracy, consistency with the proposition requirements, and validity of the question answer; if so, outputting the candidate question.

[0019] By adopting the above technical solution, the question setting system adds a preset question bank retrieval mechanism, providing an alternative solution for fast response. After receiving the question setting request, the question setting system searches for candidate questions with high matching degree in the preset question bank. If the candidate questions meet the preset output conditions, the question setting system can directly output them, which improves the response speed. This "search first, generate later" strategy not only ensures the efficiency of question generation, but also makes full use of the existing high-quality question resources to ensure the output quality.

[0020] In a second aspect, an embodiment of the present application provides a proposition system, which comprises: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code comprises computer instructions, and the one or more processors call the computer instructions to cause the proposition system to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when executed on a proposition system, enables the proposition system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a proposition system, enable the proposition system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0023] It can be understood that the proposition system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the method provided in the embodiment of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, which will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. By adopting the above technical solutions, the proposition system has established a complete proposition process through the reasoning engine and the big model. This combination ensures that the automatically generated questions not only have logical accuracy, but also have natural and fluent language expression. In addition, through multiple verification mechanisms (question grammatical accuracy, consistency of proposition requirements, and validity of question answers), the quality of the final question output is ensured, overcoming the problem that simply replacing numbers or problem situations in known math questions to generate questions may lead to logical errors, thereby improving the quality and usability of automatic proposition.

[0025] 2. By adopting the above technical solutions, the question setting system controls the three dimensions of question grammatical accuracy, question setting requirements consistency, and question answer validity, and establishes a comprehensive question quality verification mechanism. This multi-dimensional question verification method can effectively filter out unqualified questions, ensure the quality of output questions, make the question screening process more scientific and reliable, and significantly improve the practicality of the question setting system.

[0026] 3. By adopting the above technical solution, the question setting system introduces a question sorting mechanism based on difficulty characteristic parameters, so that the question difficulty can be scientifically evaluated and the questions can be output according to the difficulty level. This question difficulty evaluation method based on multi-dimensional parameters makes the generated questions more suitable for the gradient training needs in educational practice, and improves the pertinence of question application and teaching effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of a proposition method combining a large model and an inference engine in an embodiment of the present application; Figure 2 It is another flowchart diagram of the proposition method combining the large model and the reasoning engine in the embodiment of the present application; Figure 3 It is a schematic diagram of the structure of a physical device of the proposition system in the embodiment of the present application. DETAILED DESCRIPTION

[0028] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification of the present application, the singular expressions "one", "a kind of", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more of the listed items.

[0029] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0030] The following is a description of the process of the method provided by this implementation. Figure 1 , which is a flow chart of a proposition method combining a large model and an inference engine in an embodiment of the present application.

[0031] S101, receiving proposition input information, where the proposition input information includes proposition requirements; Among them, the question setting system is used to automatically generate questions; question input information refers to the information related to the question provided by the user to the question setting system; question requirements refer to the specific specification of the question, including but not limited to the question type, knowledge points to be tested, difficulty requirements, number of questions and other constraints. For example, the question requirement can be "generate a medium-difficulty word problem about a system of linear equations of two variables."

[0032] Specifically, the question setting system receives the question setting requirements input by the user through a user interface or interface. The question setting system will parse the question setting input information, extract key parameters, such as question type, knowledge point range, difficulty level, number of questions, etc., and standardize and store these key parameters to prepare for the subsequent question generation process.

[0033] S102, matching knowledge point rules whose relevance to the proposition requirement exceeds a preset relevance threshold from a preset inference rule library; Among them, the preset reasoning rule base refers to the database containing various knowledge point rules pre-stored in the proposition system; the knowledge point rule refers to the formal expression that describes specific knowledge points, including concept definitions, theorem formulas, problem-solving steps, etc.; the relevance refers to the degree of match between the knowledge point rules and the proposition requirements; the preset relevance threshold refers to the minimum matching degree requirement pre-set by the proposition system.

[0034] Specifically, first, the proposition system converts the proposition requirements into a feature vector, and then calculates the similarity between the feature vector and the feature vector of each knowledge point rule in the preset reasoning rule library. For knowledge point rules whose similarity exceeds the preset similarity threshold, the proposition system will extract them. The proposition system will consider the matching degree of multiple dimensions, including knowledge point type, difficulty level, application scenario, etc., to ensure that the selected knowledge point rules are highly relevant to the proposition requirements. If multiple qualified knowledge point rules are found, the proposition system will sort them according to the degree of relevance.

[0035] S103, combining the knowledge point rules to obtain a knowledge point combination sample question; Among them, combining the knowledge point rules means the process of integrating multiple related knowledge point rules according to logical relationships; the knowledge point combination sample question refers to the question framework obtained by combination; the logical relationship refers to the dependency, connection or deduction relationship between the knowledge point rules; the question framework represents the basic structure and components of the question.

[0036] Specifically, the proposition system will analyze the logical associations between the rules of each knowledge point and determine the feasibility of the rule combination. First, the proposition system constructs a knowledge point rule relationship diagram to identify the dependencies and combination order between knowledge point rules. Then, the proposition system will select the appropriate combination method based on the teaching rules and cognitive rules to ensure that the combined knowledge point combination sample questions have reasonable knowledge point coverage and difficulty gradient. The proposition system will also check the integrity and rationality of the combination results to ensure that the generated knowledge point combination sample questions meet the basic proposition specifications.

[0037] S104, reasoning the knowledge point combination sample question through a reasoning engine to generate a formal language expression question set, wherein the formal language expression question set includes a plurality of formal language expression questions; Among them, the reasoning engine refers to a software component used for logical reasoning and calculation; the formal language expression question refers to the question content described in a standardized and formal language; the formal language expression question set refers to a set of multiple formal language expression questions obtained through the reasoning engine.

[0038] Specifically, first, the proposition system inputs the knowledge point combination sample questions into the reasoning engine. Based on these inputs, the reasoning engine generates multiple logical question variants through formal reasoning methods such as logical deduction and constraint solving. During the reasoning process, the proposition system ensures that all generated questions meet mathematical logic and reasoning rules, and generates formal language expression questions through parameter transformation and structural adjustment. Finally, the proposition system organizes all legal reasoning results into a standard set of formal language expression questions.

[0039] Optionally, in general, reasoning on the knowledge point combination sample question through an inference engine to generate a set of formal language expression questions can be achieved in the following ways, which are not limited here: extracting key elements and logical relationships in the knowledge point combination sample question; based on the inference engine, formalizing the key elements and the logical relationships to construct a formal language grammar structure tree; traversing different branch paths of the formal language grammar structure tree to generate multiple formal language expression variants; summarizing the formal language expression variants to obtain the formal language expression question set.

[0040] S105, based on a preset translation template, converting the formal language expression topic set into a natural language expression topic set, wherein the preset translation template includes a correspondence between the formal language expression and the natural language expression, and the natural language expression topic set includes a plurality of natural language expression topics; Among them, the preset translation template refers to the set of conversion rules from formal language to natural language pre-defined by the proposition system; the correspondence between formal language expression and natural language expression is used to represent the mapping rules between formal language structure and natural language expression; natural language expression topic refers to the topic content described in the language used by humans in daily life; the natural language expression topic set refers to the set of all natural language expression topics after conversion; the conversion process refers to the process of converting formal language expression into natural language expression.

[0041] Specifically, first, the proposition system loads the preset translation template, performs a structural analysis on the formal language expression questions, and identifies the grammatical components, mathematical symbols, logical relationships and other elements. Then, the proposition system converts these elements into natural language expressions in turn according to the corresponding relationships defined in the preset translation template. During the conversion process, the proposition system will consider the fluency and expression habits of natural language, add necessary conjunctions and modal particles, adjust the word order and sentence structure, and ensure that the converted natural language expression questions are natural and smooth. At the same time, the proposition system will also maintain the accuracy and professionalism of mathematical concepts to ensure that the conversion does not change the essential meaning of the question.

[0042] S106, rewriting the natural language expression topic set by using a large model to generate an optimized topic set, wherein the optimized topic set includes a plurality of optimized topics; Among them, the big model refers to an artificial intelligence model with powerful natural language processing capabilities, such as GPT; the rewriting process refers to the process of language optimization and expression improvement of the original natural language expression question set; the optimized question refers to the question content after optimization by the big model; the optimized question set refers to the set of all rewritten optimized questions; language optimization refers to improvements in expression methods, vocabulary, tone, etc.

[0043] Specifically, first, the proposition system constructs appropriate prompt words and combines them with natural language expression questions to form an input sequence. The proposition system provides these input sequences to the big model, and uses the language understanding and generation capabilities of the big model to rewrite and optimize the natural language expression questions. During the optimization process, the big model will consider multiple aspects, including the vividness, fun, and applicability of the language, while maintaining the professionalism and accuracy of the questions. The proposition system will screen and verify the output results of the big model to ensure that the rewritten optimized questions meet educational needs while maintaining the original knowledge point requirements and difficulty level.

[0044] Optionally, under normal circumstances, the natural language expression topic set is rewritten by a large model to generate an optimized topic set, which can be achieved in the following way, which is not limited here: splicing the preset rewriting prompt words with each natural language expression topic in the natural language expression topic set to obtain multiple splicing results; inputting the splicing results into the large model to obtain the optimized topic after the natural language expression topic is rewritten; and adding the optimized topic to the optimized topic set.

[0045] S107, determining whether each optimized question in the set of optimized questions meets a preset output condition, wherein the preset output condition includes the grammatical accuracy of the question, the consistency of the question requirements, and the validity of the answer to the question; Among them, the preset output conditions refer to the question quality evaluation criteria pre-set by the question setting system; the grammatical accuracy of the question refers to the degree of standardization of the question in language expression; the consistency of the question requirements refers to the degree of conformity of the question with the original question requirements; the validity of the question answer refers to whether the question has a clear solution path and correct answer.

[0046] Specifically, first, the proposition system performs a grammatical analysis on the optimized question to check whether the words, sentence patterns, punctuation, etc. in it conform to the language specifications. Then, the proposition system calculates the similarity between the optimized question and the original proposition requirements by comparing feature vectors to ensure that the optimized question does not deviate from the original goal. Next, the proposition system calls the solution module to verify whether there is a clear solution idea and a unique correct answer for the optimized question. During the verification process, the proposition system generates a detailed evaluation report, records the test results of each dimension, and determines whether the optimized question can be output based on the preset output conditions.

[0047] Optionally, under normal circumstances, judging whether each optimized question in the set of optimized questions meets preset output conditions, and the preset output requirements include question grammatical accuracy, question requirement consistency, and question answer validity, which can be achieved in the following ways, which are not limited here: performing syntactic analysis on the optimized question to obtain a syntactic analysis result; based on the syntactic analysis result, judging whether the optimized question meets the preset question grammatical accuracy standard; calculating the similarity between the knowledge point feature vector of the optimized question and the knowledge point feature vector of the question requirement; judging whether the similarity exceeds a preset similarity threshold to determine whether the optimized question meets the preset question requirement consistency standard; solving the optimized question to obtain a solution result; based on the solution result, judging whether the optimized question meets the preset question answer validity standard.

[0048] S108: Output the optimization questions in the optimization question set that meet the preset output condition.

[0049] Among them, the optimized questions that meet the preset output conditions refer to qualified questions that have passed the quality inspection; the output process refers to the process of providing qualified questions to users.

[0050] Specifically, first, the question setting system classifies and organizes the optimized questions that meet the preset output conditions, and marks and sorts them according to characteristics such as difficulty, type, and knowledge points. The question setting system will generate complete supporting information for each optimized question that meets the preset output conditions, including detailed problem-solving steps, reference answers, knowledge point analysis, etc. During the output process, the question setting system will select the appropriate output format based on user needs, which can be a text document, structured data, or other specific format.

[0051] Optionally, under normal circumstances, outputting the optimized questions in the optimized question set that meet the preset output conditions can be achieved in the following manner, which is not limited here: extracting difficulty characteristic parameters of the optimized questions in the optimized question set that meet the preset output conditions, the difficulty characteristic parameters including the number of knowledge points, the difficulty coefficient of the knowledge points, and the number of problem-solving steps; determining the difficulty score of the optimized question based on the number of knowledge points, the difficulty coefficient of the knowledge points, and the number of problem-solving steps; and outputting the optimized questions in the optimized question set that meet the preset output conditions in order of the difficulty scores.

[0052] By adopting the above technical solutions, the proposition system has established a complete proposition process through the reasoning engine and the big model. This combination ensures that the automatically generated questions not only have logical accuracy, but also have natural and fluent language expression. In addition, through multiple verification mechanisms (question grammatical accuracy, consistency of proposition requirements, and validity of question answers), the quality of the final question output is ensured, overcoming the problem that simply replacing numbers or problem situations in known math questions to generate questions may lead to logical errors, thus improving the quality and usability of automatic proposition.

[0053] The following is a more detailed description of the process of the method provided by this implementation. Figure 2 , which is another flow chart of the proposition method combining the large model and the reasoning engine in the embodiment of the present application.

[0054] S201, receiving proposition input information, where the proposition input information includes proposition requirements; For details, please refer to step S101, which will not be described in detail here.

[0055] S202, searching the preset question bank for candidate questions whose matching degree with the question requirement exceeds a preset matching degree threshold; Among them, the preset question bank refers to the standardized question database pre-stored in the question setting system; candidate questions refer to questions that are preliminarily screened out from the preset question bank and may meet the question setting requirements; matching degree refers to the degree of conformity between the candidate questions and the question setting requirements in various dimensions; the preset matching degree threshold refers to the minimum matching degree requirement pre-set by the question setting system.

[0056] Specifically, first, the question system extracts features from the question requirements and generates feature vectors containing dimensions such as knowledge points, difficulty, and question types. Then, the question system uses efficient index structures (such as inverted indexes or vector indexes) to quickly search the preset question bank and calculate the degree of match between each question in the preset question bank and the question requirements. The question system considers the matching of multiple dimensions, including knowledge point coverage, difficulty matching, question type similarity, etc., and calculates the comprehensive matching score based on the preset weights. For questions whose matching degree exceeds the preset matching degree threshold, the question system saves them as candidate questions.

[0057] S203, determining whether the candidate question meets the preset output conditions, where the preset output requirements include question grammatical accuracy, question requirement consistency, and question answer validity; Specifically, the question setting system determines whether the candidate question meets the preset output conditions. Please refer to step S107, which will not be repeated here.

[0058] S204: If the conditions are met, output the candidate question; The output process refers to the process of providing candidate questions that meet preset output conditions to the user.

[0059] Specifically, the question setting system outputs candidate questions with preset output conditions, which can be seen in step S108 and will not be described in detail here.

[0060] S205, matching knowledge point rules whose relevance to the proposition requirement exceeds a preset relevance threshold from a preset inference rule library; For details, please refer to step S102, which will not be described in detail here.

[0061] S206, combining the knowledge point rules to obtain a knowledge point combination sample question; For details, please refer to step S103, which will not be described in detail here.

[0062] S207, reasoning the knowledge point combination sample question through a reasoning engine to generate a formal language expression question set, wherein the formal language expression question set includes a plurality of formal language expression questions; For details, please refer to step S104, which will not be described in detail here.

[0063] S208, based on a preset translation template, converting the formal language expression topic set into a natural language expression topic set, wherein the preset translation template includes a correspondence between the formal language expression and the natural language expression, and the natural language expression topic set includes a plurality of natural language expression topics; For details, please refer to step S105, which will not be described in detail here.

[0064] S209, rewriting the natural language expression topic set by using a large model to generate an optimized topic set, wherein the optimized topic set includes a plurality of optimized topics; For details, please refer to step S106, which will not be described in detail here.

[0065] S210, determining whether each optimized question in the set of optimized questions meets a preset output condition, wherein the preset output condition includes the grammatical accuracy of the question, the consistency of the question requirements, and the validity of the answer to the question; For details, please refer to step S107, which will not be described in detail here.

[0066] S211, performing error type analysis on the optimization questions that do not meet the preset output condition; Among them, error type refers to the classification of specific reasons for not meeting the preset output conditions; error type analysis refers to the process of diagnosing problems in questions; problem diagnosis criteria include multiple dimensions such as grammatical errors, knowledge point deviations, and inappropriate difficulty. For example, a question may be judged as unqualified due to inaccurate language expression, high difficulty, or unclear solution steps.

[0067] Specifically, first, the question-setting system establishes an error diagnosis framework, which includes multiple evaluation dimensions such as grammar norms, knowledge point coverage, difficulty control, and answer validity. The question-setting system will conduct multi-angle scanning of optimized questions that do not meet the preset output conditions, analyze grammatical problems through natural language processing tools, discover knowledge point deviations through knowledge graph comparison, evaluate difficulty deviations through difficulty models, and check answer problems through problem-solving verification modules. The question-setting system will mark the specific error type code for each problem found, and generate a detailed problem description report to provide a clear direction for subsequent correction work.

[0068] S212, selecting a corresponding correction prompt word from a preset correction rule library according to the error type; Among them, the preset correction rule library refers to the database of correspondence between error types and correction prompt words pre-established by the proposition system; the correction prompt words are prompt sentences used to guide the large model to correct questions. For example, for grammatically incorrect error types, the proposition system may choose "Please standardize the language expression to ensure compliance with grammatical rules" as the correction prompt word.

[0069] Specifically, first, the question setting system searches the preset correction rule library according to the error type code to find the correction template corresponding to the error type. The question setting system will parameterize the correction template according to the specific error situation and question characteristics, and generate targeted correction prompts. If there are multiple error types, the question setting system will organize multiple correction prompts in the preset priority order to ensure the orderly progress of the correction process.

[0070] S213, inputting the modified prompt word into the large model to obtain a modified optimization question; Specifically, the question-setting system constructs a standardized input sequence, combining the original question content, error analysis results, and correction prompts in a specific format. The question-setting system sets appropriate context information and constraints to ensure that the big model understands the correction requirements. After receiving the output of the big model, the question-setting system performs a preliminary verification of the correction results to check whether they meet the expected goals. If the correction results are still insufficient, the question-setting system adjusts the prompts or uses alternative correction strategies for retry.

[0071] S214: Output the optimization questions in the optimization question set that meet the preset output condition.

[0072] For details, please refer to step S108, which will not be described in detail here.

[0073] The following describes the proposition system in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of the structure of a physical device of the proposition system in an embodiment of the present application.

[0074] It should be noted that Figure 3 The structure of the proposition system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0075] like Figure 3 As shown, the proposition system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 to the random access memory (RAM) 303, such as executing the method described in the above embodiment. Various programs and data required for system operation are also stored in the RAM 303. The CPU 301, ROM 302 and RAM 303 are connected to each other through a bus 304. The input / output (I / O) interface 305 is also connected to the bus 304.

[0076] The following components are connected to the I / O interface 305: an input section 306 including an audio input device, a button switch, etc.; an output section 307 including a liquid crystal display (LCD) and an audio output device, an indicator light, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0077] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 309, and / or installed from a removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present invention are performed.

[0078] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0079] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings.

[0080] Specifically, the proposition system of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, the proposition method combining the large model and the reasoning engine provided in the above embodiment is implemented.

[0081] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the proposition system described in the above embodiment; or may exist independently without being assembled into the proposition system. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of the proposition system, the proposition system implements the proposition method combining the large model and the reasoning engine provided in the above embodiment.

[0082] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0083] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0084] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A proposition method combining a large model and an inference engine, characterized in that: Applied to a proposition system, the method comprises: Receiving proposition input information, wherein the proposition input information includes proposition requirements; Matching knowledge point rules whose relevance to the proposition requirement exceeds a preset relevance threshold from a preset inference rule library; Combining the knowledge point rules to obtain knowledge point combination sample questions; Reasoning the knowledge point combination sample questions through a reasoning engine to generate a formal language expression question set, wherein the formal language expression question set includes a plurality of formal language expression questions; Based on a preset translation template, converting the formal language expression topic set into a natural language expression topic set, wherein the preset translation template includes a correspondence between a formal language expression and a natural language expression, and the natural language expression topic set includes a plurality of natural language expression topics; Rewriting the natural language expression topic set by using a large model to generate an optimized topic set, wherein the optimized topic set includes a plurality of optimized topics; Determine whether each optimized question in the set of optimized questions meets preset output conditions, wherein the preset output requirements include question grammatical accuracy, question requirement consistency, and question answer validity; Output the optimization topics in the optimization topic set that meet the preset output conditions.

2. The method according to claim 1, characterized in that The method of using the reasoning engine to reason on the knowledge point combination sample questions to generate a set of questions expressed in a formal language specifically includes: Extracting key elements and logical relationships in the knowledge point combination sample questions; Based on the inference engine, the key elements and the logical relationships are formalized to construct a formal language syntax structure tree; Traversing different branch paths of the formal language syntax structure tree to generate multiple formal language expression variants; The formal language expression variants are aggregated to obtain the formal language expression topic set.

3. The method according to claim 1, characterized in that The rewriting of the natural language expression topic set by using the large model to generate an optimized topic set specifically includes: splicing the preset rewriting prompt word with each natural language expression topic in the natural language expression topic set to obtain multiple splicing results; Inputting the splicing result into the large model to obtain an optimized question after rewriting the natural language expression question; Add the optimization topic to the optimization topic set.

4. The method according to claim 1, characterized in that: The determining whether each optimized question in the set of optimized questions meets a preset output condition, wherein the preset output requirement includes the grammatical accuracy of the question, the consistency of the proposition requirements, and the validity of the answer to the question, specifically includes: Performing syntactic analysis on the optimization topic to obtain a syntactic analysis result; Based on the syntactic analysis result, determining whether the optimized question meets the preset question grammatical accuracy standard; Calculate the similarity between the knowledge point feature vector of the optimized question and the knowledge point feature vector of the question requirement; Determine whether the similarity exceeds a preset similarity threshold to determine whether the optimized question meets a preset question requirement consistency standard; Solving the optimization problem to obtain a solution result; Based on the solution result, determine whether the optimized question meets the preset question answer validity standard.

5. The method according to claim 1, characterized in that: After the step of determining whether each optimized question in the set of optimized questions meets preset output conditions, wherein the preset output requirements include question grammatical accuracy, question requirement consistency, and question answer validity, the method further includes: Performing error type analysis on optimization questions that do not meet the preset output conditions; According to the error type, select a corresponding correction prompt word from a preset correction rule library; The modified prompt word is input into the large model to obtain a modified optimization question.

6. The method according to claim 1, characterized in that The outputting of the optimization questions in the optimization question set that meet the preset output condition specifically includes: Extracting difficulty characteristic parameters of the optimization questions that meet the preset output condition from the optimization question set, wherein the difficulty characteristic parameters include the number of knowledge points, the difficulty coefficient of the knowledge points, and the number of problem-solving steps; Determining the difficulty score of the optimized question according to the number of knowledge points, the difficulty coefficient of the knowledge points, and the number of problem-solving steps; The optimized questions in the optimized question set that meet the preset output conditions are output in sequence according to the order of the difficulty scores.

7. The method according to claim 1, characterized in that After the step of receiving proposition input information, wherein the proposition input information includes proposition requirements, the method further includes: Retrieving candidate questions in a preset question bank whose matching degree with the proposition requirement exceeds a preset matching degree threshold; Determine whether the candidate question meets the preset output conditions, wherein the preset output requirements include question grammatical accuracy, question requirement consistency, and question answer validity; If satisfied, the candidate topic is output.

8. A proposition system, characterized in that: The proposition system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the proposition system to execute the method described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a proposition system, the proposition system is caused to execute the method as claimed in any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a proposition system, the proposition system is caused to execute the method according to any one of claims 1 to 7.

Images