Method for question analysis, electronic equipment and computer readable storage medium

Abstract

The invention discloses a method for question analysis, electronic equipment and a computer readable storage medium. The method comprises the steps of obtaining question information of a target question and historical learning data of a user; determining a target gear to which the historical learning condition of the user belongs according to the historical learning data, and determining a question analysis strategy matched with the target gear; based on the question information, the question analysis strategy and first prompt information, using a preset large language model and a retrieval enhancement generation module to generate a question analysis script; and generating a question analysis video by using a text video model based on the question analysis script, the reference video and the second prompt information. By utilizing the scheme disclosed by the invention, personalized question analysis can be realized, and adaptive explanation contents are provided for students with different knowledge levels.

Description

Technical Field

[0001] The present disclosure generally relates to the field of artificial intelligence technology. More specifically, the present disclosure relates to a method, electronic device, and computer-readable storage medium for question parsing. Background Art

[0002] With the vigorous development of artificial intelligence technology, its application in the field of education is becoming increasingly widespread. With its powerful data processing and analysis capabilities, artificial intelligence can provide diverse support for learning (such as regular learning and competitive learning). During the learning process, the intelligent learning system can analyze the student's learning status based on the student's learning data, such as answer status and learning time, customize personalized learning plans for students, recommend appropriate learning materials and exercises, and improve learning efficiency. In terms of teaching, teachers use artificial intelligence-assisted teaching tools to more accurately grasp students' knowledge weaknesses, provide targeted tutoring, and optimize teaching content and methods. At the same time, artificial intelligence also plays an important role in the examination and assessment process, enabling intelligent marking and score analysis, and providing students and teachers with detailed learning feedback reports.

[0003] However, the current application of AI in exam assessment still faces numerous challenges. For example, the analysis of questions after practice tests lacks personalization. The standardized text-based analysis fails to fully account for the differences in comprehension and knowledge mastery among students. As a result, some students struggle to fully understand the solution even after reading the answer key, failing to effectively address their knowledge gaps.

[0004] In view of this, there is an urgent need to provide a solution for question analysis so as to achieve personalized question analysis and provide adapted explanation content for students with different knowledge levels. Summary of the Invention

[0005] In order to at least solve one or more of the technical problems mentioned above, the present disclosure proposes solutions for question parsing in the following aspects.

[0006] In a first aspect, the present disclosure provides a method for error parsing, comprising: obtaining topic information of a target topic and historical learning data of a user; determining a target level to which the user's historical learning situation belongs based on the historical learning data, and determining a topic parsing strategy that matches the target level; generating a topic parsing script based on the topic information, the topic parsing strategy, and a first prompt information using a preset large language model and a retrieval enhancement generation module; and generating a topic parsing video based on the topic parsing script, a reference video, and a second prompt information using a Vincent video model.

[0007] In some embodiments, the question information includes question content and at least one knowledge point associated with the question, the question content includes the question stem, answer, difficulty and / or options; the historical learning data includes previous test scores; the question analysis strategy includes analysis method and exercise difficulty.

[0008] In some embodiments, determining the target gear to which the user's historical learning situation belongs based on the historical learning data includes: determining the user's historical average score based on the previous test scores; and matching the historical average score with the score range corresponding to the preset gear to determine the target gear.

[0009] In some embodiments, after generating a question analysis video, the method also includes displaying the question analysis video; and after displaying the question analysis video, the method also includes: upon receiving a question input by the user about the question analysis video, generating question and answer information using the preset large language model based on the question, the question information, the question analysis script, the question analysis strategy and the third prompt information; and displaying the question and answer information.

[0010] In some embodiments, after displaying the question-and-answer information, the method further includes: displaying inquiry information, wherein the inquiry information is used to ask the user whether he or she has any questions about the question analysis video; upon receiving confirmation information input by the user, obtaining at least one practice question based on the question information and the question analysis strategy, wherein the confirmation information indicates that the user has no questions about the question analysis video; and displaying the at least one practice question.

[0011] In some embodiments, before displaying the inquiry information, the method further includes: if no user input is received within a preset time period, using the preset large language model to generate the inquiry information; and after displaying the at least one practice question, the method further includes: receiving the answer content input by the user; determining the user's practice score based on the answer content; and determining whether to continue to obtain practice questions based on the practice score.

[0012] In some embodiments, obtaining practice questions based on the question information and the question parsing strategy includes: matching the knowledge points in the question information with the knowledge points in the question bank to obtain all questions under the same knowledge points; matching the practice difficulty in the question parsing strategy with the question difficulty of all questions to obtain questions whose question difficulty matches the practice difficulty as the practice questions.

[0013] In some embodiments, the preset gears include a first gear, a second gear, a third gear and a fourth gear; and determining the question parsing strategy that matches the target gear includes: when the target gear is the first gear, the parsing method in the question parsing strategy that matches the target gear is extended parsing, and the practice difficulty is competition-level difficulty, wherein the extended parsing is to parse at least one knowledge point in the question information under different application scenarios and different knowledge fields; when the target gear is the second gear, the parsing method in the question parsing strategy that matches the target gear is multiple problem-solving methods analysis, and the practice difficulty is lower than the competition-level difficulty and higher than the difficulty in the question information; when the target gear is the third gear, the parsing method in the question parsing strategy that matches the target gear is step-by-step parsing, and the practice difficulty is the difficulty in the question information; when the target gear is the fourth gear, the parsing method in the question parsing strategy that matches the target gear is step-by-step parsing and example parsing, and the practice difficulty is lower than the difficulty in the question information.

[0014] In a second aspect, the present disclosure provides an electronic device comprising: a processor; and a memory storing program instructions for question parsing, wherein when the program instructions are executed by the processor, the method described in the first aspect and its multiple embodiments are implemented.

[0015] In a third aspect, the present disclosure provides a computer-readable storage medium having stored thereon program instructions for question parsing, wherein when the program instructions are executed by a processor, the method described in the first aspect and its multiple embodiments are implemented.

[0016] The solution for question analysis, as described above, utilizes both question information and the user's historical learning data to categorize the user's learning history into different levels and match them with corresponding question analysis strategies. This shifts from the traditional standardized text analysis model, fully accounting for differences in user knowledge and comprehension, providing tailored explanations for students of varying knowledge levels and enabling personalized question analysis.

[0017] The solution for question parsing provided by this disclosure uses a preset large language model and a retrieval enhancement generation module to generate a question parsing script, and then uses Wensheng video to generate a question parsing video. This multimodal presentation method is more attractive and intuitive than the traditional single text analysis. The video format can show the problem-solving process in a richer way, such as through animation demonstrations, step-by-step decomposition, etc., to help students better understand complex knowledge points, thereby improving learning effects. Moreover, personalized analysis content can enable students to understand the questions faster, reduce the time spent on understanding the problem-solving ideas, and thus improve learning efficiency.

[0018] Furthermore, when generating a question parsing script, the search enhancement generation module retrieves relevant information, so that the parsing content is not limited to the current question, but can also be associated with other related knowledge points and questions. This helps students build a more complete knowledge system and deepen their understanding and memory of knowledge points. For example, when parsing a math problem, it may be associated with similar question types or different application scenarios of related theorems learned previously, allowing students to understand knowledge from multiple perspectives, strengthen their mastery of knowledge, and improve their ability to apply knowledge in learning. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present disclosure are shown in an illustrative and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0020] Figure 1 is an exemplary flow chart illustrating a method for question parsing according to some embodiments of the present disclosure;

[0021] Figure 2 is an exemplary schematic diagram showing the association relationship between topics and knowledge points according to the present disclosure;

[0022] Figure 3 is an exemplary schematic diagram illustrating a process of generating a topic parsing script using a preset large language model and a retrieval enhancement generation module according to the present disclosure;

[0023] Figure 4 is an exemplary flow chart illustrating a method for question parsing according to other embodiments of the present disclosure;

[0024] Figure 5 is an exemplary schematic diagram illustrating obtaining practice questions according to the present disclosure;

[0025] Figure 6 is an exemplary flow chart illustrating a method for question parsing implemented based on a learning system according to an embodiment of the present disclosure;

[0026] Figure 7 is a diagram illustrating an exemplary interaction process of a client according to an embodiment of the present disclosure;

[0027] Figure 8 FIG. 4 is a block diagram showing an exemplary structure of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0028] The following will clearly and completely describe the technical solutions in the embodiments of this disclosure in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this disclosure, not all of them. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this disclosure.

[0029] It should be understood that when the terms "first," "second," "third," and "fourth" are used in the claims, specification, and drawings of this application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprising" used in the specification and claims of this disclosure indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or combinations thereof.

[0030] It should also be understood that the terminology used in this disclosure is for the purpose of describing specific embodiments only and is not intended to limit the disclosure. As used in this disclosure and the claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should be further understood that the term "and / or" as used in this disclosure and the claims refers to any and all possible combinations of one or more of the associated listed items, including and including these combinations.

[0031] As used in this specification and claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0032] The specific embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0033] Figure 1 The following is an exemplary flow chart of a method 100 for question parsing according to an embodiment of the present disclosure. It is understood that method 100 can be executed by any appropriate device with data processing capabilities, including but not limited to terminal devices, processors, and servers. Terminal devices include but are not limited to smartphones, smart learning machines, personal computers, laptops, tablet computers, and portable wearable devices.

[0034] like Figure 1As shown, at step S101, method 100 can obtain the topic information of the target topic and the user's historical learning data. At step S102, method 100 can determine the target gear to which the user's historical learning situation belongs based on the historical learning data, and determine the topic parsing strategy that matches the target gear. Then, at step S103, method 100 can generate a topic parsing script based on the topic information, the topic parsing strategy and the first prompt information using a preset large language model and a retrieval enhancement generation module. At step S104, method 100 can generate a topic parsing video based on the topic parsing script, the reference video and the second prompt information using a Vincent video model. Finally, at step S105, method 100 can display the topic parsing video.

[0035] In the actual learning process, not only do incorrectly answered questions need to be analyzed, but some correctly answered questions also need to be explained. For example, in a test, multiple-choice questions or true-or-false questions may still be answered correctly even if the user does not fully understand the question. Another example is a test question, even if the user does not fully understand all the knowledge points behind the question, there is still a possibility of answering correctly. In response to these situations, in addition to generating a question analysis video when a user makes an incorrect answer, the present disclosure also allows users to independently mark questions that need to be analyzed, giving users more learning autonomy, enabling them to actively participate in the learning process, and conduct targeted learning based on their actual situation.

[0036] Based on this, the aforementioned target question can be a question with an incorrect answer, or a question with a preset mark, which is used to indicate that the user (such as a student participating in the study) has a need to solve the question. It is understandable that those skilled in the art may actually need to select a specific expression form of the preset mark, and this disclosure does not limit this.

[0037] In the disclosed embodiment, the question information of any question, including the target question, may include the question content and at least one knowledge point associated with the question. The question content may include the question stem, answer, and difficulty. In addition, if the question type is a multiple-choice question, the question content may also include options.

[0038] In actual application, there is a many-to-many relationship between questions and knowledge points, that is, one question can involve multiple knowledge points, and at the same time, one knowledge point can be tested by multiple questions. Figure 2 To understand, as Figure 2 As shown, question 1 may involve three knowledge points, namely, knowledge point A, knowledge point B, and knowledge point C, and knowledge D may be tested by questions 2, 3, and 4.

[0039] In the disclosed embodiments, a user's historical learning data may include all previous exam times, all previous exam scores, all previous exam answering times, and all previous exam answering accuracy rates. In step S102, the target gear for the user's historical learning situation is determined based on the historical learning data. The following operations may be performed: determining the user's historical average score based on all previous exam scores; and matching the historical average score with the score range corresponding to the preset gear to determine the target gear. In one embodiment, the preset gears may include a first gear, a second gear, a third gear, and a fourth gear, each corresponding to a different score range.

[0040] In one implementation scenario, the score range corresponding to the first gear may be [85, 100], indicating that the user's learning situation is excellent; the score range corresponding to the second gear may be [70, 85], indicating that the user's learning situation is good; the score range corresponding to the third gear may be [60, 70], indicating that the user's learning situation is average; and the score range corresponding to the fourth gear may be (60, 0], indicating that the user's learning situation is poor. It can be understood that the preset gears and score ranges shown here are merely exemplary and illustrative. In order to achieve refined management of user learning, in actual applications, more gears and score ranges can be divided. The implementation method and implementation entity of the present disclosure are not limited thereto, but can be changed without deviating from the spirit of the present disclosure.

[0041] In the disclosed embodiment, the question parsing strategy may include a parsing method and a difficulty of practice. In order to achieve personalized question parsing, a plurality of question parsing strategies may be pre-configured, and different preset gears may match different question parsing strategies. In the case where the aforementioned preset gears include a first gear, a second gear, a third gear, and a fourth gear, a total of four gears, the question parsing strategy may also include a first question parsing strategy, a second question parsing strategy, a third question parsing strategy, and a fourth question parsing strategy, a total of four, and the parsing methods and difficulty of practice in the four question parsing strategies are different from each other.

[0042] Specifically, when the target gear is the first gear, the question parsing strategy that matches the target gear is the first question parsing strategy. In the first question parsing strategy, the parsing method is extended parsing, and the practice difficulty is competition-level difficulty. Extended parsing here refers to parsing at least one knowledge point in the question information under different application scenarios and different knowledge fields. Competition-level difficulty usually refers to the complexity and challenge of the questions in a specific field (such as mathematics, physics, informatics, chemistry, biology, etc.) reaching the standard of competition level. In other words, the analysis of the first question parsing strategy focuses on the in-depth analysis of difficult questions, emphasizing the expansion and application of knowledge points. By introducing real competition questions or innovative question types involving knowledge points in the questions, and providing high-level problem-solving ideas and techniques, it can further enhance the user's ability.

[0043] When the target gear is the second gear, the question parsing strategy that matches the target gear is the second question parsing strategy. In the second question parsing strategy, the parsing method is to use multiple problem-solving methods to analyze and practice questions with a difficulty lower than the aforementioned competition-level difficulty and higher than the difficulty in the question information of the target question. In other words, the analysis focus of the second question parsing strategy is to consolidate medium-difficulty questions and fill in knowledge gaps. By analyzing the knowledge points involved in the questions in detail and providing multiple problem-solving methods, users can flexibly apply the knowledge points. In addition, by appropriately introducing high-difficulty questions with the same knowledge points, the user's ability can be gradually improved.

[0044] When the target gear is the third gear, the matching question analysis strategy is the third question analysis strategy. In this strategy, the analysis method is step-by-step analysis, and the exercise difficulty is the difficulty specified in the question information. In other words, the third question analysis strategy focuses on consolidating the foundation and resolving errors in easy and medium difficulty questions. By providing detailed step-by-step analysis, users are ensured to understand each step. In addition, practice questions with the same knowledge points and the same difficulty level can be added to strengthen the user's memory of the knowledge points covered in the questions.

[0045] When the target gear is the fourth gear, the question parsing strategy that matches the target gear is the fourth question parsing strategy. In the fourth question parsing strategy, the parsing method is step-by-step parsing and example parsing, and the difficulty of the exercises is lower than the difficulty in the question information. In other words, the analysis of the fourth question parsing strategy focuses on starting from the basics and solving the mistakes in simple questions. The knowledge points involved in the question are explained in detail step by step, and easy-to-understand language and examples are used to reduce the difficulty of understanding. In addition, by providing a large number of basic question exercises with the same knowledge points and lower difficulty, users can gradually improve.

[0046] In the aforementioned step S103, the preset large language model can be any large language model (LLM) that has appeared or may appear in the future. As long as the large language model can process text data, perform logical reasoning and generate natural language text, it can be used to implement the solution for question parsing disclosed in this disclosure. Figure 3 Understand the process of generating question parsing scripts using a preset large language model and retrieval enhancement generation module.

[0047] like Figure 3 As shown, the question information and question parsing strategy will be input as input data into the preset large language model, and the preset large language model will pass the question information and question parsing strategy to the retrieval enhancement generation module. The retrieval enhancement generation module has a question bank that stores a large amount of question information, which includes the question content of simulation questions and real questions (including question stems, answers and difficulty), at least one knowledge point associated with the question, etc. After receiving the question information and question parsing strategy input, the retrieval enhancement generation module will use a specific retrieval algorithm to search in the question bank based on the question information and question parsing strategy. For example, if the target question is a knowledge point about mathematical functions, and the user's question parsing strategy is the second question parsing strategy, the retrieval enhancement generation module will give priority to retrieving question information related to the function knowledge point and of moderate difficulty. These question information may include solution ideas for similar questions, analysis of common errors, etc.

[0048] Afterward, the search enhancement generation module returns the retrieved question information to the pre-set large language model. Combined with the first prompt information, the large language model uses this question information to generate and output a question parsing script. For example, the large language model will reference similar question parsings retrieved, combine the specifics of the target question with the user's question parsing strategy, and generate detailed solution ideas, steps, and knowledge points tailored to the user, ultimately creating a high-quality question parsing script.

[0049] The first prompt is used to guide the pre-set large language model to generate a question parsing script that meets the requirements. In practice, the first prompt can take the following form: Please generate a question parsing script for the user based on the following question information and user-specific question parsing strategy. Question information: []; User-specific question parsing strategy: []; Other requirements: [].

[0050] As an example, the first prompt message may be: Please generate a question parsing script for the user based on the following question information and the user's unique question parsing strategy. Question information: [Question 'In triangle ABC, AB=3, BC=4, ∠ABC=60°, find the length of AC', answer Difficulty 'medium', related knowledge point 'Law of Cosines']; User-specific question analysis strategy: [Second question analysis strategy, analysis method is to analyze with multiple problem-solving methods, and the difficulty of the exercise is lower than the aforementioned competition-level difficulty and higher than the difficulty in the question information of the target question]; Other requirements: [The question analysis script first explains the concept of the cosine theorem, then uses two different methods to solve the problem, each method is explained in detail step by step, and finally summarizes the key points of the problem-solving].

[0051] In the aforementioned step S104, the Vincent video model can be any Vincent video model that has appeared currently or may appear in the future. As long as the Vincent video model can understand the input information (such as the content of the question analysis script, the visual information and style characteristics conveyed by the reference video, etc.), generate logical video content, and support multimodal fusion, it can be used to implement the solution for question analysis disclosed herein. Figure 3 Understand the process of using the Wensheng video model to generate question analysis videos. Figure 3 As shown, the question parsing script, reference video, and second prompt information are input into the Vincent video model as input data. After receiving the input data, the Vincent video model can generate and output the question parsing video based on the question parsing script and reference video under the guidance of the second prompt information.

[0052] The aforementioned reference videos provide the Vincent video model with key reference information in terms of visuals and style. The Vincent video model can generate a question analysis video that meets specific style and quality requirements based on elements such as the reference video's layout, color matching, and shot switching method, combined with the question analysis script. In actual applications, reference videos can cover a variety of forms and styles to meet the learning needs and preferences of different users. For example, the form of the reference video can include animation demonstrations, real-life shooting, expert explanations, and other explanation forms, and the style of the reference video can include different explanation styles such as concise style, detailed style, and humorous style. In this way, when generating a user-specific question analysis video, it is possible to select a suitable reference video based on the user's characteristics and needs, thereby improving the applicability and attractiveness of the video.

[0053] The aforementioned second prompt is used to guide the Wensheng video model to generate a problem-solving video that meets the requirements. As an example, the second prompt can be described in the following form: Please generate a 3-5 minute exclusive problem-solving video based on the provided problem-solving script [specific script content, such as introducing the problem at the beginning, explaining the concept of the cosine theorem, two problem-solving methods, etc.] and the reference video [video style, screen switching method, and other related reference information]. The explanation in the video should be clear and the speed should be moderate. The pictures should be concise and clear. When explaining each problem-solving step, the corresponding formula and graphics should be displayed synchronously on the screen, and the key content should be marked with eye-catching colors.

[0054] Combination of the above Figure 1 A method 100 for question parsing is described. The method 100 obtains the question information of the target question and the user's historical learning data, divides the user's historical learning situation into different levels, and matches the corresponding question parsing strategy. This changes the previous standardized text parsing model, fully considers the differences in knowledge mastery and comprehension ability among different users, and can provide adapted explanation content for students with different knowledge levels, thus realizing personalized question parsing. Based on this, for users with better learning conditions (such as high-scoring levels), more in-depth and expansive analysis content can be provided to help them further improve their abilities; for users with relatively weaker learning conditions, starting from the basic knowledge points, the problem-solving ideas are explained in detail, so that students with different knowledge levels can get adapted explanation content and effectively make up for their respective knowledge gaps.

[0055] In practice, users may have various questions after watching a problem-solving video. However, the existing answer-solving model lacks a feedback mechanism, preventing users from receiving timely answers to their questions. This lack of interactive Q&A leads to a continuous accumulation of problems encountered during the learning process, reducing learning efficiency. Furthermore, the consolidation of learning is weak, as the existing answer-solving model lacks an effective consolidation mechanism. Even if users have a vague understanding of the solution, it is difficult to strengthen their understanding, leading to knowledge bottlenecks during problem-solving, which in turn affects the overall learning effect and quality.

[0056] Based on this, we will then combine Figure 4 The method 400 for question parsing provided by other embodiments of the present disclosure is described. Figure 4 As shown. At step S401, the method 400 can obtain the title information of the target title and the user's historical learning data. At step S402, the method 400 can determine the target gear to which the user's historical learning situation belongs based on the historical learning data, and determine the title parsing strategy that matches the target gear. At step S403, the method 400 can generate a title parsing script based on the title information, the title parsing strategy and the first prompt information using a preset large language model and a retrieval enhancement generation module. At step S404, the method 400 can generate a wrong title parsing video based on the title parsing script, the reference video and the second prompt information using a Vincent video model. At step S405, the method 400 can display the title parsing video. It should be noted that, in this embodiment, the contents of steps S401 to S405 are consistent with steps S101 to S105 in the previous embodiment, and will not be repeated here.

[0057] Next, at step S406, upon receiving a user's input regarding a question analysis video, a pre-set large language model can be used to generate a question-answering message based on the question, question information, question analysis script, question analysis strategy, and third prompt information. At step S407, the question-answering message can be displayed. The question-answering message provides an answer to the user's question, thus establishing an instant feedback mechanism that ensures timely answers to user questions and ensures that the user truly grasps the knowledge points covered in the target question.

[0058] The aforementioned third prompt information is used to guide the preset large voice model to generate question-and-answer information that meets the requirements. As an example, the third prompt information can be described in the following form: You are a professional learning tutoring assistant, and now a user has raised a question about a question. Please answer based on the following information: question analysis script [script content containing detailed analysis logic], question information [question stem, answer, difficulty, related knowledge points, etc.]; user-specific question analysis strategy [such as the third question analysis strategy, the analysis method is step-by-step analysis, and the practice difficulty is the difficulty in the question information]; user questions [such as in the second problem-solving method, when taking the square root in the last step, why only positive values ​​are taken?]. Please answer the user's questions in detail and patiently, and provide a simple exercise with similar knowledge points after the answer.

[0059] In some implementation scenarios, after studying the aforementioned Q&A information, if the user still has questions about the problem-solving video and / or the displayed Q&A information, they can continue to enter questions. Thus, steps S406 and S407 can be repeatedly executed to generate Q&A information for the user's questions until the user's questions are completely resolved, achieving seamless interactive learning.

[0060] In other implementation scenarios, after displaying the question-answering information, method 400 may continue to execute steps S408 to S411. Specifically, at step S408, if no user input is received within the preset time length, the aforementioned preset large language model may be used to generate an inquiry message, which is used to ask the user whether he has any questions about the question analysis video and / or the displayed question-answering information. At step S409, the inquiry message may be displayed. It is understandable that those skilled in the art may select a specific value for the preset time length according to actual needs, and this disclosure does not make any specific limitations on this.

[0061] After the inquiry information is displayed, at step S410, upon receiving a confirmation message from the user, at least one practice question may be retrieved based on the question information and the question resolution strategy. The confirmation message indicates that the user has no questions regarding the question resolution video and / or the displayed question-and-answer information. Furthermore, at step S411, at least one practice question may be displayed.

[0062] At the aforementioned step S410, the specific execution may be Figure 5 The following operations are shown to obtain practice questions based on question information and question parsing strategy: Match the knowledge points in the question information with the knowledge points in the question bank to obtain all questions under the same knowledge points (i.e. Figure 5 The problem 1 to problem N shown in the figure); Match the difficulty of the exercises in the problem analysis strategy with the difficulty of all the problems to obtain problems that match the difficulty of the exercises as the practice problems (i.e. Figure 5 Question 1 and Question 2 are shown in ).

[0063] In an embodiment of the present disclosure, after displaying at least one exercise question, the answer content input by the user can also be received, and the user's exercise score can be determined based on the answer content. Thereafter, it can be determined whether it is necessary to continue to obtain exercise questions based on the user's exercise score. In one embodiment, when the user's exercise score reaches a preset score, it can be determined that the user has mastered the knowledge points involved in the wrong questions and does not need to continue to obtain exercise questions. On the contrary, when the user's exercise score does not reach the preset score, it can be determined that the user has not mastered the knowledge points involved in the target question and it is necessary to continue to obtain exercise questions to provide them to the user for continued practice. The specific value of the preset score here can be selected by those skilled in the art according to actual needs, and this disclosure does not make specific restrictions on this.

[0064] This intensive and consolidating practice approach ensures that users truly master the knowledge points covered in the target questions, thereby comprehensively improving overall learning effectiveness and quality. Furthermore, through steps S401 to S411, a complete closed loop of question analysis, user Q&A, and practice consolidation is achieved. This not only addresses the issues of insufficient personalization, delayed feedback, and lack of consolidation in existing answer analysis models, but also significantly improves user learning efficiency and effectiveness.

[0065] From the previous description, we can see that Figures 1 to 5 All steps of the method for question parsing described above are performed by a single device, which undoubtedly increases the computing pressure of the device. In order to balance the pressure on the device, in actual applications, some steps that directly interact with the user can be deployed on the client side, and other steps can be deployed on the server side to form a learning system. Figure 6 The method 600 for question parsing based on the learning system is described in detail. Figure 6 As shown, at step S601, the user can complete the answers to the simulation test questions through the client, and the client obtains the user's answers and transmits them to the server, so that the server can judge the user's answers and provide personalized question analysis.

[0066] Correspondingly, after receiving the answer content, the server can execute the following steps S602-S605 to generate a personalized question analysis video. Specifically, at step S602, the server judges the user's answer content. If it detects that the user has answered incorrectly, it can determine the incorrectly answered question as the target question and obtain the question information of the target question and the user's historical learning data; at step S603, the server can determine the target level of the user's historical learning situation based on the historical learning data, and determine the question analysis strategy that matches the target level; at step S604, the server can generate a question analysis script based on the question information, question analysis strategy and the first prompt information using a preset large language model and a retrieval enhancement generation module; at step S605, the server can generate a question analysis video based on the question analysis script, the reference video and the second prompt information using a cultural video model.

[0067] Next, the server can execute step S606 to send the question analysis video to the user's client. Correspondingly, the client can execute step S607 to display the question analysis video so that the user can deepen his understanding of the knowledge points by watching the question analysis video. Thereafter, if the user has any questions about the question analysis video, he can send the questions to the server through the client. Accordingly, the server can generate question-answering information when receiving the questions input by the user, and send the question-answering information to the client for display. Additionally or optionally, the server can also obtain at least one practice question when it is determined that the user has no questions, and send them to the client for display so that the user can perform targeted practice.

[0068] Combination of the above Figure 6 The method 600 for question parsing implemented based on the learning system is described. It is understood that the description of each embodiment in this disclosure focuses on the differences between the embodiments, and the same or corresponding parts can be referenced to each other. For the purpose of brevity, this disclosure will not go into details one by one.

[0069] Next, combine Figure 7 The interactive process of the client according to the embodiment of the present disclosure is exemplarily introduced. Figure 7 As shown in a in Figure 1, for questions that need to be analyzed, the client can display the question, the correct answer, and the answer analysis in three areas. In the answer analysis area, a button for opening the question analysis video can be displayed, such as the "View My Exclusive Analysis" button shown in a. After the user clicks the "View My Exclusive Analysis" button, the client's display screen is updated to the following: Figure 7 As shown in b. Figure 7As can be seen from b, the client has started a dialogue between the analysis assistant and the user. In this dialogue, the analysis assistant can not only show the user the question analysis video, but also receive the user's questions and answer them until the user has no more questions. Then, the display screen of the client is updated to the following Figure 7 As shown in c. Figure 7 As can be seen from c in the figure, the client will display a prompt message indicating that the analysis is complete, such as "The dialogue is over and the analysis is complete" shown in c, and will also display buttons for opening exercises, such as Exercise 1, Exercise 2, and Exercise 3 shown in c. The user clicks any button to open the exercise corresponding to the button, and the client's display screen will also be updated to the following: Figure 7 As shown in d, the specific content of the exercise questions is displayed.

[0070] Combination of the above Figure 7 The interactive process of the client according to the embodiment of the present disclosure is described. It should be understood that the display screen or display content here is only exemplary, and the implementation method and implementation entity of the present application are not limited thereto, but can be changed without departing from the spirit of the present application.

[0071] Next, combine Figure 8 An electronic device 800 provided by an embodiment of the present disclosure is exemplarily introduced. Figure 8 As shown, the electronic device 800 according to the embodiment of the present disclosure may include a processor 801 , a memory 802 and a communication bus 803 .

[0072] In a specific embodiment, the processor 801 may be at least one of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a CPU, a controller, a microcontroller, and a microprocessor. It is understood that for different devices, the electronic device used to implement the above-mentioned processor functions may also be other electronic devices, which is not specifically limited in this embodiment.

[0073] In the embodiment of the present disclosure, the communication bus 803 is used to realize the connection and communication between the processor 801 and the memory 802; the memory 802 stores program instructions for parsing the question; when the processor 801 executes the program instructions stored in the memory 802, the present disclosure is realized. Figures 1 to 6 The described method for question parsing.

[0074] Combination of the above Figure 8 An electronic device for topic parsing that can be used to perform the present disclosure is described. It should be understood that the device structure or architecture here is merely exemplary, and the implementation method and implementation entity of the present application are not limited thereto, but can be changed without departing from the spirit of the present application. It is understandable that the description of each embodiment of the present disclosure focuses on the differences between the various embodiments, and the same or corresponding parts can be referenced to each other. For the purpose of brevity, the present disclosure will not go into details one by one.

[0075] According to the above description in combination with the accompanying drawings, those skilled in the art can also understand that the embodiments of the present application can also be implemented by software programs. Therefore, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores program instructions for wrong question analysis, and the program instructions can be used to implement the present disclosure in combination with the software program. Figures 1 to 6 The described method for error question resolution.

[0076] It should be noted that although the operations of the present method are described in a particular order in the accompanying drawings, this does not require or imply that the operations must be performed in that particular order, or that all of the operations shown must be performed to achieve the desired results. Rather, the steps depicted in the flowcharts may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into a single step, and / or a single step may be broken down into multiple steps.

[0077] Although a plurality of embodiments of the present disclosure have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art may conceive of many modifications, changes, and alternatives without departing from the ideas and spirit of the present disclosure. It should be understood that in practicing the present disclosure, various alternatives to the embodiments of the present disclosure described herein may be adopted. The appended claims are intended to define the scope of protection of the present disclosure and therefore cover equivalents or alternatives within the scope of these claims.

Claims

1. A method for question parsing, comprising: Obtain the target topic information and the user's historical learning data; Determine the target level to which the user's historical learning situation belongs based on the historical learning data, and determine a question parsing strategy that matches the target level; Based on the topic information, the topic parsing strategy and the first prompt information, a preset large language model and a search enhancement generation module are used to generate a topic parsing script; as well as Based on the question analysis script, the reference video and the second prompt information, a question analysis video is generated using the Wensheng video model.

2. The method according to claim 1, wherein The question information includes the question content and at least one knowledge point associated with the question, the question content includes the question stem, answer, difficulty and / or options; the historical learning data includes previous test scores; the question analysis strategy includes the analysis method and exercise difficulty.

3. The method according to claim 2, wherein: Determining the target gear to which the user's historical learning situation belongs based on the historical learning data includes: Determine the user's historical average score based on the previous test scores; and The historical average score is matched with a score range corresponding to a preset gear to determine the target gear.

4. The method according to claim 2, further comprising, after generating the question analysis video, displaying the question analysis video; After displaying the question analysis video, the method further includes: Upon receiving a question input by the user regarding the question analysis video, generating answer information using the preset large language model based on the question, the question information, the question analysis script, the question analysis strategy, and third prompt information; as well as The question-and-answer information is displayed.

5. The method according to claim 4, further comprising: after displaying the question-answering information; Displaying inquiry information, wherein the inquiry information is used to ask the user whether he has any questions about the question analysis video; Upon receiving confirmation information input by the user, obtaining at least one practice question according to the question information and the question analysis strategy, wherein the confirmation information indicates that the user has no questions about the question analysis video; as well as The at least one practice question is displayed.

6. The method according to claim 5, wherein: Before displaying the query information, the method further includes: generating the query information using the preset large language model if no user input is received within a preset time period; After presenting the at least one exercise question, the method further comprises: Receiving the answer content input by the user; Determining the user's practice score based on the answer content; and Determine whether to continue obtaining practice questions according to the practice results.

7. The method according to claim 5, wherein: Acquiring practice questions according to the question information and the question parsing strategy includes: Matching the knowledge points in the question information with the knowledge points in the question bank to obtain all questions with the same knowledge points; The practice difficulty in the question parsing strategy is matched with the question difficulty of all questions to obtain questions whose question difficulty matches the practice difficulty as the practice questions.

8. The method according to claim 3, wherein: The preset gears include a first gear, a second gear, a third gear, and a fourth gear; and determining a question parsing strategy that matches the target gear includes: When the target gear is the first gear, the parsing method in the question parsing strategy that matches the target gear is extended parsing and the practice difficulty is competition-level difficulty, wherein the extended parsing is to parse at least one knowledge point in the question information under different application scenarios and different knowledge fields; When the target gear is the second gear, the problem solving strategy that matches the target gear is a problem solving method that uses multiple problem solving methods, and the difficulty of the practice is lower than the competition level and higher than the difficulty in the problem information; When the target gear is the third gear, the problem solving strategy that matches the target gear is a step-by-step analysis method, and the practice difficulty is the difficulty in the problem information; When the target gear is the fourth gear, the parsing method in the question parsing strategy that matches the target gear is step-by-step parsing and example parsing, and the difficulty of the exercise is lower than the difficulty in the question information.

9. An electronic device comprising: processor; and a memory storing program instructions for question parsing, wherein when the program instructions are executed by the processor, the electronic device executes the method according to any one of claims 1-8.

10. A computer-readable storage medium storing program instructions for question parsing, wherein when the program instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images