Search

CN-122020589-A - Personalized topic exercise scheme generation method for intelligent wrong topic integration

CN122020589ACN 122020589 ACN122020589 ACN 122020589ACN-122020589-A

Abstract

The invention discloses a method for generating an individualized question exercise scheme for intelligent wrong question integration, which relates to the technical field of individualized education and comprises the steps of receiving a user wrong question record set, analyzing a question stem, options and answers, mapping knowledge points to a multi-level knowledge system tree, calculating the occurrence frequency of the knowledge points and counting wrong answer option distribution modes of associated wrong questions, determining a weak knowledge node set and a typical wrong understanding mode corresponding to the weak knowledge node set, searching matched correction exercise questions in a question bank based on the weak knowledge node set, and performing difficulty adaptation adjustment to generate a primary individualized exercise question sequence. The invention identifies the deep cognition error region by analyzing the error answer distribution mode and utilizes the knowledge system tree to locate the structural weakness, thereby generating a personalized exercise scheme capable of accurately correcting the error understanding and consolidating the knowledge system.

Inventors

  • WANG CHUANJUN
  • ZHAN DEHOU
  • HU KONGWANG

Assignees

  • 上海贝乘教育科技有限公司

Dates

Publication Date
20260512
Application Date
20260128

Claims (10)

  1. 1. The method for generating the personalized topic exercise scheme for intelligent wrong topic integration is characterized by comprising the following steps of: Receiving a wrong question record set submitted by a user through a terminal, wherein the wrong question record set comprises a user identifier, wrong question original content, a corresponding knowledge point label and a wrong answer option when answering a wrong question; carrying out structural analysis on the wrong question record set, extracting a question stem text, an option text and an answer text in the wrong question original content, and mapping the knowledge point label into a preset multi-level knowledge system tree; Calculating the occurrence frequency of each knowledge point label in the wrong question record set based on the multi-level knowledge system tree, and counting the distribution mode of wrong answer options in the wrong questions associated with each knowledge point label; Determining a weak knowledge node set reflected by the wrong question record set and a typical error understanding mode corresponding to each weak knowledge node according to the occurrence frequency and the distribution mode; Searching correction exercise questions matched with the weak knowledge node set in a preset question bank resource pool according to the weak knowledge node set and the typical error understanding mode; And performing difficulty adaptation adjustment on the retrieved correction exercise questions to generate a primary personalized exercise question sequence.
  2. 2. The method for generating a personalized topic exercise scheme for intelligent integration of wrong topic according to claim 1, wherein the method is characterized in that the wrong topic record set is subjected to structural analysis, the stem text, the option text and the answer text in the wrong topic original content are extracted, and the knowledge point label is mapped into a preset multi-level knowledge system tree, specifically: Reading the original content of each wrong question from the wrong question record set, wherein the original content of each wrong question comprises but is not limited to a plain text format and a rich text format containing format marks; Identifying a stem part and an answer option part in the wrong original content by adopting a text segmentation rule in natural language processing, wherein the text segmentation rule is based on keywords, punctuation marks and format characteristics; Adopting regular expression matching or a sequence labeling model based on deep learning to separate each independent option text from the answer option part, and identifying an answer text marked as a correct answer; extracting keywords and identifying entities from the extracted topic stem text to generate abstract feature vectors for characterizing topic topics; Matching the similarity between the abstract feature vector and an inverted index which is built in advance and based on a subject knowledge structure, and supplementing or correcting a corresponding knowledge point label for the stem text; The multi-level knowledge system tree comprises a root node, a chapter node and knowledge point leaf nodes, and each knowledge point leaf node is provided with a unique path code; using the knowledge point labels in each wrong question record as query keys, performing matching or fuzzy matching in a multi-level knowledge system tree, and positioning to the corresponding knowledge point leaf nodes or upper chapter nodes thereof; and (3) taking the path code of the successfully matched knowledge point leaf node or chapter node as the standardized knowledge point identifier of the wrong question record, and writing the standardized knowledge point identifier into the extension information of the wrong question record set.
  3. 3. The method for generating a personalized topic exercise scheme for intelligent integration of wrong topic according to claim 1, wherein the calculating the occurrence frequency of each knowledge point label in the wrong topic record set based on the multi-level knowledge system tree, and counting the distribution pattern of wrong answer options in the wrong topic associated with each knowledge point label specifically comprises: traversing the error question record set, and reading standardized knowledge point identifiers in each error question record, namely path codes in the multi-level knowledge system tree; counting the occurrence times of each unique path code in the error question record set, and dividing the occurrence times by the total number of the error question record set to obtain the relative error frequency of the corresponding knowledge point of each path code; For the error question subset with the same path code, further analyzing the error answer options of each error question record subordinate to the error question subset; taking the text content of the wrong answer option or the option identifier after standardized processing as a basic unit for analysis; counting the occurrence times of each basic unit in the wrong question subset to form a frequency distribution table of wrong answer options; Analyzing the characteristics of the frequency distribution table according to the topic type, wherein the characteristics comprise whether the characteristics are concentrated on a specific interference item, uniformly distributed on a plurality of interference items, and whether certain interference items are never selected; and the relative error frequency and the frequency distribution characteristics of the error answer options are used together as quantization index data for describing the grasping condition of the user on the corresponding knowledge points of the path codes.
  4. 4. The method for generating a personalized topic exercise scheme for intelligent integration of wrong topic according to claim 1, wherein determining the weak knowledge node set reflected by the wrong topic record set and the typical error understanding mode corresponding to each weak knowledge node according to the occurrence frequency and the distribution mode specifically comprises: Setting a threshold value of relative error frequency, and primarily marking a knowledge point corresponding to a path code with the relative error frequency exceeding the threshold value as a suspected weak knowledge node; Aiming at each suspected weak knowledge node, analyzing a corresponding wrong answer option frequency distribution table; if the wrong answer option frequency distribution table shows that the errors are highly concentrated on a specific interference item, judging that a typical wrong understanding mode aiming at the concept represented by the specific interference item exists for the user, wherein the mode is marked as 'concept confusion'; If the error answer option frequency distribution table shows that errors are uniformly distributed in a plurality of interference items and the error rate exceeds a preset threshold, judging that the overall understanding of the knowledge points by a user is fuzzy or knowledge holes exist, and marking the mode as 'understanding fuzzy'; if the wrong answer option frequency distribution table shows wrong option dispersion but the non-answer rate exceeds a preset first threshold value or the average answer time length is lower than a preset second threshold value, judging that the user has a problem solving skill or time distribution problem, wherein the mode is marked as 'strategy missing'; combining the suspected weak knowledge nodes which have close association and show similar error understanding modes according to the levels and association relations of the suspected weak knowledge nodes in the multi-level knowledge system tree to form a comprehensive weak knowledge node; And forming a binary group by each confirmed weak knowledge node and the corresponding typical error understanding mode thereof, wherein the set of the binary group is the weak knowledge node set and the typical error understanding mode corresponding to each weak knowledge node.
  5. 5. The method for generating a personalized topic exercise scheme for intelligent wrong topic integration according to claim 1, wherein the searching of the corrected exercise topic matched with the weak knowledge node set and the typical wrong understanding mode in the preset topic library resource pool specifically comprises the following steps: extracting a binary group from the weak knowledge node set, wherein the binary group comprises path codes and typical error understanding modes of the weak knowledge nodes; searching all topics marked with the same or similar path codes in the topic knowledge point index of the topic library resource pool according to the path codes to form an initial candidate topic set; filtering the initial candidate topic set according to the typical misunderstanding pattern; If the typical misunderstanding mode is "concept confusion", then preferentially selecting those topics or options that explicitly relate to the topics identified by the mismatching concepts identified by the mismatching options frequency distribution, or setting the topics for the mismatching concepts by the interference item; if the typical misunderstanding mode is "understanding fuzzy", the questions of the knowledge point basic definition and the core principle are preferentially selected to be examined, or the question combination of the same knowledge point is examined from multiple angles; If the typical misunderstanding mode is 'strategy missing', the topics with typical aspects in the problem solving thought or the problems needing specific problem solving skills are preferentially selected; Calculating the topic relevance scores of the filtered topics with the weak knowledge nodes and typical misunderstanding modes; sorting the topics according to the topic relevance scores, and selecting a plurality of topics with top ranking as correction exercise topics matched with the binary groups; and repeating the searching process for each binary group in the weak knowledge node set to finally obtain the correction exercise question set matched with all weak knowledge nodes and the error understanding mode.
  6. 6. The method for generating a personalized exercise program for intelligent wrong question integration according to claim 1, wherein the method is characterized in that the method comprises the steps of performing difficulty adaptation adjustment on the retrieved corrected exercise questions to generate a preliminary personalized exercise question sequence, and the method is specifically as follows: acquiring historical capability level evaluation data of the user, wherein the historical capability level evaluation data comprises the comprehensive scores of the accuracy rate and the answering time of past exercises of the user on similar knowledge points; Reading a preset reference difficulty coefficient from a question library resource pool for each question in the retrieved correction exercise question set; Establishing a difficulty adjustment function, wherein the input of the difficulty adjustment function is the historical capability level evaluation data of a user and a reference difficulty coefficient of a question, and the input is an adjusted difficulty expected value applicable to the user; the difficulty adjustment function follows the logic that for users with historical capability level evaluation data greater than or equal to a preset capability threshold, the adjusted difficulty expected value floats on the basis of a reference difficulty coefficient; screening the topics in the correction exercise topic collection according to the calculated adjusted difficulty expected value, and removing topics with the adjusted difficulty expected value exceeding the acceptable range of the current capability of the user; Grouping the selected topics according to the associated weak knowledge nodes; Inside each group, arranging according to the adjusted difficulty expected value of the title from low to high to form a progressive structure from easy to difficult; sequencing all the groups according to the relative error frequency of the weak knowledge nodes corresponding to the groups, and arraying the topic groups corresponding to the weak knowledge nodes with high relative error frequency in front; All the sequenced question groups are sequentially connected to form a linear primary personalized exercise question sequence which is increased according to the importance of the weak points and then the in order question difficulty.
  7. 7. The method for generating a personalized topic exercise scheme for intelligent wrong topic integration according to claim 1, further comprising the steps of obtaining a historical exercise performance record of a user, and optimizing the number and sequence of the preliminary personalized exercise topic sequences according to the historical exercise performance record, wherein the number and sequence of the preliminary personalized exercise topic sequences are as follows: acquiring a history exercise performance record of the user, wherein the record comprises average accuracy rate, average response time and the proportion of giving up answers of the user to different difficulties and different types of questions; Analyzing the historical exercise performance record, and fitting an exercise fatigue curve of the user, wherein the exercise fatigue curve describes the trend of the change of the accuracy rate along with the increase of the number of exercise questions in the continuous exercise process of the user; Determining a recommended topic number threshold value which is not lower than a preset correct rate threshold value and an attention level threshold value and can be maintained by a user in single effective exercise according to the exercise fatigue curve; Comparing the total number of topics of the preliminary personalized exercise topic sequence with the suggested topic number threshold; If the total number of the topics exceeds the threshold value of the number of the recommended topics, deleting part of the topics from the end of the sequence according to the sequence from low to high relative error frequency of the associated weak knowledge nodes until the total number of the topics is equal to the threshold value of the number of the recommended topics; if the total number of the questions does not exceed the threshold value of the number of the recommended questions, keeping the sequence unchanged; in the sequence after the quantity adjustment, checking whether a plurality of continuous topics belong to the same topic type; If yes, judging whether the continuous multi-channel questions meet a preset type fatigue triggering condition according to the average response time and the giving up response proportion of the user to the question types of the continuous multi-channel questions in the history exercise performance record; If the fatigue of the type is caused, on the premise of ensuring the increasing order of the internal difficulty of the same weak knowledge node question group, the questions of the same type are properly arranged at intervals, or a small number of questions of other types are inserted to be used as buffering; And after the number and the sequence are adjusted, an optimized personalized exercise question sequence is formed.
  8. 8. The method for generating a personalized exercise program for intelligent wrong question integration according to claim 1, wherein the method further comprises the steps of packaging the optimized personalized exercise program sequence with corresponding weak knowledge node labels and correction target specifications to generate an executable personalized exercise program data packet, and the method is characterized in that: Generating a unique exercise item identifier for each topic in the optimized personalized exercise topic sequence; Associating, for each exercise item identifier, information including complete topic content, standard answers and resolution of the topic, weak knowledge node path coding for the topic, and short correction target specification words generated according to a typical misunderstanding pattern; The correction target description text is generated according to the typical error understanding mode template and is used for prompting the key points of the exercise of the subject to the user; Organizing the logic sequence of the optimized personalized exercise question sequence, each exercise item identifier and all associated information thereof according to a preset data exchange format; The data exchange format defines a storage structure of scheme header information, a practice entry list, and a scheme metadata portion; recording the generation time, the target user identifier and the expected completion time of the exercise scheme in scheme header information; storing the complete data of each exercise item in sequence in an exercise item list; in scheme metadata, recording weak knowledge node summaries and main correction target summaries covered by the scheme; The organized data is serialized into a separate executable personalized exercise program package containing all necessary information and instructions.
  9. 9. The method for generating a personalized topic exercise program for intelligent wrong topic integration according to claim 1, wherein the method further comprises pushing the personalized exercise program data packet to a user terminal and starting exercise process monitoring, specifically: Transmitting the executable personalized exercise scheme data packet to a user terminal application program bound with a user identifier through a message queue or a network interface; after receiving the executable personalized exercise scheme data packet, the user terminal application program analyzes the data packet and sequentially presents exercise questions on a user interface; When a user starts to answer a first question, the user terminal application program sends an exercise start signal to a server, the server records an exercise start time stamp, and a monitoring timer for the exercise is started; the user terminal application program records the following behavior data of the user on each question, namely selected answer options, a history record of modifying the answer, the stay time of each question page and the final time of submitting the answer; the user terminal application program transmits the behavior data back to the server in real time or near real time; The server receives and stores the behavior data and associates the behavior data with the corresponding exercise item identifiers and the question standard answers; the server calculates elapsed exercise time according to the current time and the exercise start time stamp, and compares the elapsed exercise time with the estimated completion time in the scheme header information; If the stay time of the user on a certain topic is found to be abnormal and exceeds the historical average reaction time of the topic to which the topic belongs, or the whole exercise progress is seriously delayed from the predicted time line, the server can generate prompt information to gently remind the user of the attention time through the user terminal application program.
  10. 10. The method for generating a personalized topic exercise scheme with intelligent integration of wrong topic according to claim 1, wherein the method further comprises the steps of collecting answer results and answer time of each topic in the personalized exercise scheme data packet by a user in an exercise process, and updating a wrong topic record set and a historical exercise performance record of the user, specifically: After the user finishes or terminates all the questions in the personalized exercise scheme data packet, the server integrates the final answer results of all the questions collected from the user terminal application program; Comparing the final answer result with the standard answer of each question, and marking the questions with wrong answers; Extracting exercise item identifiers of questions with wrong answers, and backtracking associated question contents, wrong answers submitted by users and corresponding weak knowledge node path codes; Adding the information obtained by backtracking into a wrong question record set of the user as a new wrong question record, and simultaneously recording the context identification of the exercise; Recording the answering time of the user on all questions, regardless of errors; The answer result, answer time, question type and difficulty expected value after the question is adjusted of each question are used as a new exercise performance record and updated into the historical exercise performance record of the user; According to the updated wrong question record set and the history practice performance record, recalculating the grasping degree change trend of the user on the relevant knowledge points; The updated wrong question record set becomes input data when the next round of personalized question exercise scheme generation method is executed, and exercise feedback and scheme optimization closed loop is realized.

Description

Personalized topic exercise scheme generation method for intelligent wrong topic integration Technical Field The invention belongs to the technical field of personalized education, and particularly relates to a personalized topic exercise scheme generation method for intelligent wrong topic integration. Background In the current personalized learning system, the exercise generation of the wrong questions mainly depends on the frequency statistics of knowledge points related to the wrong questions. The system judges weak points according to the wrong question number threshold value of the single knowledge point, and extracts questions of the same knowledge point from the question bank for repeated training. This approach attributes the user's mistakes to a poor grasp of isolated knowledge points, whose technical logic is limited to quantity-based mechanical matching. The prior art solutions have drawbacks. The choice of wrong answer from the user cannot be analyzed in a patterned way, so that the different cognitive causes behind the error cannot be distinguished. The method has the advantages that the generated exercise is lack of pertinence, deep misunderstanding of a user is difficult to effectively intervene, only discrete knowledge points are focused, and the knowledge points are not placed in a preset multi-level knowledge system for association analysis. The weak links of users often appear as systematic defects on knowledge chains or clusters, but the existing method cannot identify the weak point set with inherent structural association, so that exercise topics are scattered, and it is difficult to help users build complete and coherent knowledge structures. What is needed is a mistopic processing method that can parse the wrong answer pattern to identify typical misunderstandings and rely on a structured knowledge system to locate a set of weak knowledge nodes of relevance. The method is based on analysis of wrong answer distribution patterns and structural calculation of a multi-level knowledge system, and realizes transition from correcting surface errors to correcting wrong cognition patterns, and from training isolated knowledge points to consolidating the knowledge system. Disclosure of Invention The present invention aims to solve at least one of the technical problems existing in the prior art; therefore, the invention provides a personalized topic exercise scheme generation method for intelligent wrong topic integration, which comprises the following steps: Receiving a wrong question record set submitted by a user through a terminal, wherein the wrong question record set comprises a user identifier, wrong question original content, a corresponding knowledge point label and a wrong answer option when answering a wrong question; carrying out structural analysis on the wrong question record set, extracting a question stem text, an option text and an answer text in the wrong question original content, and mapping the knowledge point label into a preset multi-level knowledge system tree; Calculating the occurrence frequency of each knowledge point label in the wrong question record set based on the multi-level knowledge system tree, and counting the distribution mode of wrong answer options in the wrong questions associated with each knowledge point label; Determining a weak knowledge node set reflected by the wrong question record set and a typical error understanding mode corresponding to each weak knowledge node according to the occurrence frequency and the distribution mode; Searching correction exercise questions matched with the weak knowledge node set in a preset question bank resource pool according to the weak knowledge node set and the typical error understanding mode; And performing difficulty adaptation adjustment on the retrieved correction exercise questions to generate a primary personalized exercise question sequence. Further, the structuralized analysis is performed on the wrong question record set, the question stem text, the option text and the answer text in the wrong question original content are extracted, and the knowledge point label is mapped into a preset multi-level knowledge system tree, specifically: Reading the original content of each wrong question from the wrong question record set, wherein the original content of each wrong question comprises but is not limited to a plain text format and a rich text format containing format marks; Identifying a stem part and an answer option part in the wrong original content by adopting a text segmentation rule in natural language processing, wherein the text segmentation rule is based on keywords, punctuation marks and format characteristics; Adopting regular expression matching or a sequence labeling model based on deep learning to separate each independent option text from the answer option part, and identifying an answer text marked as a correct answer; extracting keywords and identifying entities from the extracted topic s