Preparation path recommendation method, device, equipment, medium and system

Abstract

The application provides a test preparation path recommendation method and device, electronic equipment, computer readable storage medium and system. When a test preparation path recommendation period is reached, a test preparation portrait record, a current test preparation stage record, a target test range record and a test point dependency graph are obtained. A test point priority portrait is generated, pre-preparation readiness verification and stage adaptation are performed, and a stage test preparation path sequence is formed. A question type matching record is generated through semantic alignment of question semantic records and test point semantic records, and then a test preparation path recommendation record is generated, and the test preparation portrait record is updated according to the execution feedback.

Description

Technical Field

[0001] This application relates to the field of intelligent education data processing technology, and more specifically, to a method, apparatus, electronic device, computer-readable storage medium, and system for recommending test preparation paths. Background Technology

[0002] As online education platforms, intelligent question bank systems, and learning analysis systems are increasingly used in middle school exam preparation, a large amount of process data is continuously generated, including learning behaviors, answer feedback, review time, and knowledge mastery status. How to transform this process data into an executable exam preparation plan, and how to adapt this plan to the learning objectives of different exam preparation stages, is a key technical problem that intelligent exam preparation systems need to address.

[0003] Existing test preparation recommendation schemes typically generate learning tasks based on a fixed review plan or a single learning indicator. These schemes generally first obtain students' recent answer results, then determine the content to be reviewed based on the number of incorrect answers, the degree of weakness in knowledge points, or the teacher's pre-set review order, and then retrieve relevant questions from a question bank and push them to students. While this type of scheme can provide students with review content to some extent, its processing logic usually leans towards a single sorting or fixed template, and the recommendation results have a weak perception of changes in the test preparation process, the connection between knowledge bases, and semantic differences in questions.

[0004] However, the above-mentioned solutions still have significant technical shortcomings. Because the generation of preparation tasks is based on a relatively singular basis, there is a tendency for sequential jumps between different test points, resulting in low coherence of the preparation path. Furthermore, the system fails to adequately verify the mastery of relevant basic test points before assigning advanced test points, which can easily lead to a mismatch between subsequent learning tasks and existing knowledge. In addition, the continued use of similar recommendation logic after changes in the preparation stage makes it difficult to adapt to the different requirements of coverage, task length, and review pace at different stages. When question matching relies primarily on keywords in the question stem or question type tags, semantic discrepancies between questions and target test points can easily occur, thereby reducing the executability of the recommended preparation path records and the effectiveness of subsequent updates. Summary of the Invention

[0005] This application provides a method, apparatus, electronic device, computer-readable storage medium, and system for recommending test preparation paths, in order to at least alleviate the aforementioned technical problems.

[0006] A method for recommending test preparation paths includes: When the recommended preparation path cycle is reached, obtain the preparation profile record, current preparation stage record, target exam scope record, and test point dependency graph corresponding to the target exam scope record of the candidate. Based on the test preparation profile records, test point priority status identifiers are set for candidate test points in the test point dependency graph to generate test point priority profiles. Based on the test point dependency graph, a pre-readiness check is performed on the test point priority profile, and pre-correction entries are generated for candidate test points that do not meet the pre-readiness conditions. Based on the current preparation stage records, the priority profiles of test points and the pre-correction items are adapted to the stage to generate a stage preparation path sequence. Based on the sequence of preparation paths, obtain the semantic records of questions, align the semantic records of questions with the semantic records of test points, and generate question type matching records. Based on the sequence of preparation paths and the matching records of question types, a preparation path recommendation record is generated, and the preparation profile record is updated based on the execution feedback of the preparation path recommendation record.

[0007] Optionally, the system obtains the candidate's preparation profile record, current preparation stage record, target exam scope record, and test point dependency graph corresponding to the target exam scope record, including: Read the learning behavior records of the test takers from the test preparation system and convert the learning behavior records into behavior attention tags; Read the test-taking feedback records of the test takers during the question-answering process from the test preparation system, and convert the feedback records into error reason markers; Read the mastery status records of test takers in the test point dimension from the test preparation system, and convert the mastery status records into mastery gap markers; Read the historical review time records of the test takers in the test point dimension from the test preparation system, and convert the historical review time records into review interval markers; Combine behavioral attention markers, error cause markers, mastery gap markers, and review interval markers to create a test preparation profile record; The current preparation stage is determined based on the preparation calendar status corresponding to the recommended preparation path cycle. The target exam scope is determined based on the target exam configuration corresponding to the preparation target. The exam point dependency graph is then obtained based on the target exam scope record.

[0008] Optionally, based on the test preparation profile records, test point priority status identifiers are set for candidate test points in the test point dependency graph, generating a test point priority profile, including: In the test point dependency graph, candidate test points are determined based on the target test scope record; The behavioral attention markers, error cause markers, mastery gap markers, and review interval markers in the test preparation profile are mapped to candidate test points to obtain test point status mapping records corresponding to the candidate test points. Based on the test site status mapping record, generate the test site priority status identifier corresponding to the candidate test site, and generate the test site priority profile based on the candidate test site carrying the test site priority status identifier.

[0009] Optionally, based on the test point dependency graph, a pre-readiness check is performed on the test point priority profile, and pre-correction entries are generated for candidate test points that do not meet the pre-readiness conditions, including: Read the candidate test points carrying the test point priority status identifier from the test point priority profile, and obtain the preceding test point chain corresponding to the candidate test point from the test point dependency graph; Based on the test preparation profile record, determine the mastery status of each pre-test point in the pre-test point chain, and perform pre-test point readiness verification on the pre-test point chain according to the preset pre-test readiness conditions. If there is a candidate test point in the current test point chain that does not meet the prerequisite readiness conditions, set a downgrade flag for the candidate test point and identify the candidate test point that does not meet the prerequisite readiness conditions as a prerequisite reinforcement item. The candidate test point is generated based on a combination of the downgrade identifier and the preceding reinforcement items.

[0010] Optionally, prerequisite reinforcement items participate in the generation of the stage preparation path sequence, including: After generating the prerequisite reinforcement items, determine whether there are upstream prerequisite test points in the prerequisite reinforcement items; If it exists, the corresponding upstream prerequisite test point chain is obtained according to the test point dependency graph, and the mastery status of each upstream prerequisite test point in the upstream prerequisite test point chain is determined according to the test preparation profile record. Based on the prerequisite readiness conditions, perform prerequisite readiness checks on the upstream prerequisite test point chain, and update the prerequisite reinforcement entries according to the check results. Set the pre-readiness verification completed pre-compliance entries to an insertable state, so that pre-compliance entries carrying the insertable state are arranged before the corresponding candidate test points during stage adaptation.

[0011] Optionally, based on the records of the current preparation stage, the priority profile of test points and the pre-exam correction items are adapted to the current stage to generate a sequence of preparation paths for each stage, including: Extract the phase task boundaries, test point coverage strategies, and path length control identifiers from the current preparation phase records; Based on the phase task boundaries, candidate test points for each phase are selected from the test point priority profile; The test site priority status identifier is obtained by adjusting the test site priority status identifier of the candidate test sites in the stage according to the test site coverage strategy. Based on the path length control identifier, a portion of the candidate test points carrying the stage priority status identifier are extracted, and a stage preparation path sequence is generated according to the order in which the pre-correction reinforcement entries in the pre-correction entries are located before the corresponding candidate test points, and the demotion identifiers in the pre-correction entries are applied to the corresponding candidate test points.

[0012] Optionally, the current preparation stage record is generated based on the preparation calendar status; when the preparation calendar status indicates a stage switch, the stage task boundary, test point coverage strategy and path length control identifier in the current preparation stage record are updated according to the switched preparation stage. When the next preparation path recommendation cycle arrives, the test point priority profile and pre-correction items will be re-adapted to the current preparation stage based on the updated current preparation stage record, so that the newly generated stage preparation path sequence matches the preparation stage after the switch.

[0013] Optionally, based on the stage preparation path sequence, question semantic records are obtained, and the question semantic records are semantically aligned with the test point semantic records to generate question type matching records, including: Obtain the corresponding semantic records of test points based on the test point identifiers in the phased preparation path sequence; Retrieve candidate questions that match the test point semantic records from the preset question bank index, and extract the question semantic records corresponding to the candidate questions; Semantic anchor point alignment is performed between the semantic records of the questions and the semantic records of the test points to obtain the question matching identifier; Based on the candidate questions carrying question matching identifiers, a question type matching record for the corresponding test point is generated.

[0014] Optionally, a test preparation path recommendation record is generated based on the test preparation path sequence and question type matching record, and the test preparation profile record is updated based on the execution feedback of the test preparation path recommendation record, including: Establish path node associations for candidate questions in the test point and question type matching records in the phased preparation path sequence; Based on the relationship between path nodes, a test preparation path recommendation record is generated. Among them, the stage test preparation path sequence determines the order of test points in the test preparation path recommendation record, and the question type matching record determines the candidate questions corresponding to each test point. Obtain the path execution feedback generated after the recommended test preparation path is executed, update the behavior attention mark, error cause mark, mastery gap mark and review interval mark based on the path execution feedback, and write the updated marks into the test preparation profile record for use in the next test preparation path recommendation cycle.

[0015] A test preparation path recommendation device, comprising: The periodic data acquisition module is used to acquire the preparation profile record, current preparation stage record, target exam scope record, and test point dependency graph corresponding to the target exam scope record when the recommended preparation path period is reached. The priority profile generation module is used to set the priority status identifier of candidate test points in the test point dependency graph based on the test preparation profile records, and generate test point priority profiles. The pre-correction generation module is used to perform pre-readiness verification on the priority profile of test points based on the test point dependency graph, and generate pre-correction entries for candidate test points that do not meet the pre-readiness conditions. The stage path generation module is used to adapt the test point priority profile and the pre-correction items to the current test preparation stage records, and generate a stage test preparation path sequence. The question type matching and generation module is used to obtain the question semantic records according to the stage preparation path sequence, semantically align the question semantic records with the test point semantic records, and generate question type matching records. The recommendation feedback update module is used to generate a preparation path recommendation record based on the stage preparation path sequence and question type matching record, and update the preparation profile record based on the execution feedback of the preparation path recommendation record.

[0016] An electronic device includes a processor and a memory, wherein the memory stores computer instructions, and the processor reads and executes the computer instructions to implement the above-mentioned recommended study path method.

[0017] A computer-readable storage medium storing computer instructions that, when executed by a processor, implement the above-mentioned method for recommending study paths.

[0018] A test preparation path recommendation system includes a front-end test preparation learning device and a cloud-based test preparation path recommendation platform; The front-end exam preparation learning device is used to: collect exam preparation process data from the exam candidates, send the exam preparation process data to the cloud-based exam preparation path recommendation platform, and display the exam preparation path recommendation records returned by the cloud-based exam preparation path recommendation platform; the front-end exam preparation learning device is also used to collect path execution feedback generated after the exam candidates execute the exam preparation path recommendation records, and send the path execution feedback to the cloud-based exam preparation path recommendation platform; The cloud-based exam preparation path recommendation platform is used for: maintaining exam preparation profile records based on exam preparation process data; maintaining records for the current exam preparation stage based on the exam preparation path recommendation cycle; and maintaining a test point dependency graph based on the target exam scope record. It generates test point priority profiles based on the exam preparation profile records and test point dependency graphs; performs pre-readiness checks on the test point priority profiles based on the test point dependency graphs, generates pre-correction items, and performs stage adaptation on the test point priority profiles and pre-correction items based on the current exam preparation stage records, generating a stage-specific exam preparation path sequence. It performs semantic alignment between question semantic records and test point semantic records based on the stage-specific exam preparation path sequence, generates question type matching records, and generates exam preparation path recommendation records based on the stage-specific exam preparation path sequence and question type matching records. Finally, it updates the exam preparation profile records based on path execution feedback, enabling the updated exam preparation profile records to participate in the next exam preparation path recommendation cycle.

[0019] Optionally, the exam preparation process data collected by the front-end exam preparation learning device includes learning behavior records, answer feedback records, mastery status records, and historical review time records; the cloud-based exam preparation path recommendation platform maintains exam preparation profile records based on the exam preparation process data, including: converting learning behavior records into behavior attention tags, answer feedback records into error reason pointing tags, mastery status records into mastery gap tags, historical review time records into review interval tags, and combining behavior attention tags, error reason pointing tags, mastery gap tags, and review interval tags into exam preparation profile records.

[0020] Optionally, the cloud-based exam preparation path recommendation platform maintains the current exam preparation stage record according to the exam preparation path recommendation cycle, including: determining the current exam preparation stage based on the exam preparation calendar status corresponding to the exam preparation path recommendation cycle, and writing the stage task boundaries, test point coverage strategies, and path length control identifiers corresponding to the current exam preparation stage into the current exam preparation stage record; when the exam preparation calendar status indicates a stage switch, updating the stage task boundaries, test point coverage strategies, and path length control identifiers in the current exam preparation stage record according to the switched exam preparation stage.

[0021] Optionally, the cloud-based exam preparation path recommendation platform maintains a test point dependency graph based on the target exam scope record, including: determining candidate test points within the target exam scope based on the target exam scope record, and writing the candidate test points into the test point dependency graph; writing the preceding test point chain corresponding to the candidate test point into the test point dependency graph based on the preceding dependency relationship between test points within the target exam scope, and writing the corresponding upstream preceding test point chain when the preceding test point in the preceding test point chain has an upstream preceding test point.

[0022] Optionally, the cloud-based exam preparation path recommendation platform generates exam point priority profiles based on exam preparation profile records and exam point dependency graphs. This includes: determining candidate exam points in the exam point dependency graph based on the target exam scope record; mapping behavioral attention markers, error cause pointing markers, mastery gap markers, and review interval markers in the exam preparation profile records to the candidate exam points, obtaining exam point status mapping records corresponding to the candidate exam points; generating exam point priority status identifiers corresponding to the candidate exam points based on the exam point status mapping records; and generating exam point priority profiles based on the candidate exam points carrying exam point priority status identifiers.

[0023] Optionally, the cloud-based test preparation path recommendation platform performs pre-readiness verification on the test point priority profile based on the test point dependency graph, and generates pre-correction entries, including: reading candidate test points carrying test point priority status identifiers from the test point priority profile, and obtaining the corresponding pre-test point chain from the test point dependency graph; determining the mastery status of each pre-test point in the pre-test point chain based on the test preparation profile records, and performing pre-readiness verification on the pre-test point chain based on preset pre-readiness conditions; when there are pre-test points in the pre-test point chain that do not meet the pre-readiness conditions, setting a downgrade identifier for the candidate test point, and identifying the pre-test point that does not meet the pre-readiness conditions as a pre-reinforcement entry; and generating the pre-correction entry corresponding to the candidate test point based on the combination of the downgrade identifier and the pre-reinforcement entry.

[0024] Optionally, the cloud-based exam preparation path recommendation platform adapts the exam point priority profile and pre-exam correction items to the current exam preparation stage records, generating a stage exam preparation path sequence. This includes: obtaining the stage task boundary, exam point coverage strategy, and path length control identifier from the current exam preparation stage records; filtering candidate exam points from the exam point priority profile based on the stage task boundary; adjusting the exam point priority status identifier of the candidate exam points based on the exam point coverage strategy to obtain the stage priority status identifier; and extracting a portion of the exam points from the candidate exam points carrying the stage priority status identifier based on the path length control identifier, and generating the stage exam preparation path sequence according to the order in which the pre-exam reinforcement items in the pre-exam correction items are located before the corresponding candidate exam points, and the downgrade identifiers in the pre-exam correction items are applied to the corresponding candidate exam points.

[0025] Optionally, the cloud-based exam preparation path recommendation platform performs semantic alignment between question semantic records and exam point semantic records based on the phased exam preparation path sequence to generate question type matching records. This includes: obtaining the corresponding exam point semantic records based on the exam point identifiers in the phased exam preparation path sequence; obtaining candidate questions that match the exam point semantic records from a preset question bank index and extracting the question semantic records corresponding to the candidate questions; aligning the question semantic records and exam point semantic records with semantic anchors to obtain question matching identifiers; and generating question type matching records for the corresponding exam points based on the candidate questions carrying question matching identifiers.

[0026] Optionally, the cloud-based exam preparation path recommendation platform generates exam preparation path recommendation records based on the phased exam preparation path sequence and question type matching records, including: establishing path node associations between the test points in the phased exam preparation path sequence and the candidate questions in the question type matching records; generating exam preparation path recommendation records based on the path node associations, wherein the phased exam preparation path sequence determines the order of test points in the exam preparation path recommendation records, and the question type matching records determine the candidate questions corresponding to each test point; and sending the exam preparation path recommendation records to the front-end exam preparation and learning devices.

[0027] Optionally, the cloud-based exam preparation path recommendation platform updates the exam preparation profile record based on path execution feedback, including: extracting the exam preparation subject's execution status of the exam points and candidate questions in the exam preparation path recommendation record from the path execution feedback; updating the behavior attention marker, error cause marker, mastery gap marker, and review interval marker based on the execution status, and writing the updated behavior attention marker, error cause marker, mastery gap marker, and review interval marker into the exam preparation profile record.

[0028] The technical advantages of the technical solution provided in this application are: To address the issues of single-criteria ranking and fragmented study paths in existing technologies, this application simultaneously acquires study profile records, current study stage records, target exam scope records, and exam point dependency graphs upon reaching the recommended study path period. This allows study path recommendations to move beyond relying solely on single answer results or fixed review sequences. Instead, it uses study profile records as the basis for individual status and the target exam scope records and their corresponding exam point dependency graphs to limit the sources of candidate exam points. Furthermore, based on study profile records, priority status identifiers are set for candidate exam points in the exam point dependency graphs, and priority profiles are generated, enabling candidate exam points to express their priority status to the study subject before recommendation. Compared to traditional solutions that only push review tasks based on the number of incorrect answers, question type tags, or fixed schedules, this application uses study profile records and exam point dependency graphs to jointly constrain the priority status of candidate exam points. This makes the order of exam points in the recommended study path records more accurately reflect the correspondence between the study subject's current status and the target exam scope, thereby mitigating the fragmentation problem caused by disordered jumps between different exam points.

[0029] To address the issue of insufficient verification of the prerequisite status of high-level test points in existing technologies, this application does not directly convert test point priority profiles into recommended study paths. Instead, it performs prerequisite readiness verification on the test point priority profiles based on a test point dependency graph and generates prerequisite correction entries for candidate test points that do not meet the prerequisite readiness conditions. Through this process, the test point dependency graph is not merely used to provide a display of the test point range or hierarchy, but directly participates in the verification process before candidate test points enter the stage study path sequence; the prerequisite correction entries are used to correct the path structure of candidate test points that do not meet the prerequisite readiness conditions. Compared to the traditional recommendation scheme where high-level test points may directly enter the recommendation sequence, this application can technically constrain the prerequisite status of candidate test points before generating the stage study path sequence, making the recommended study path records more consistent with the learning succession relationships between test points, thereby reducing path execution difficulties caused by insufficient prerequisites.

[0030] To address the issue of low adaptability of recommendation logic to changes in the preparation stage in existing technologies, this application, after pre-readiness verification, adapts the test point priority profile and pre-correction items based on the current preparation stage record to generate a stage-specific preparation path sequence. This process ensures that the current preparation stage record not only exists as display stage information but also participates in the path-based processing of the test point priority profile and pre-correction items; the stage-specific preparation path sequence is formed jointly by the test point priority profile, pre-correction items, and the current preparation stage record. Compared to traditional solutions that use similar recommendation logic throughout the preparation process, this application can adjust the path formation method based on the current preparation stage record within different preparation path recommendation cycles. This makes the stage-specific preparation path sequence more aligned with the task boundaries, coverage requirements, and path length control needs of the preparation stage, thereby improving the adaptability of the preparation path recommendation record to changes in the preparation process.

[0031] To address the issue of semantic mismatch between questions and target test points in existing technologies, this application obtains question semantic records based on the stage-based exam preparation path sequence and semantically aligns these records with test point semantic records to generate question type matching records. This process ensures that question matching does not solely rely on question stem keywords or question type tags, but rather establishes a semantic correspondence between test points in the stage-based exam preparation path sequence and both the question and test point semantic records. This allows the question type matching records to align with the test points in the stage-based exam preparation path sequence. Compared to traditional keyword-based question retrieval methods, this application reduces the impact of differences in question wording, synonyms, or implicit conditions on test point matching through semantic alignment. This results in a higher semantic correlation between candidate questions and corresponding test points in the exam preparation path recommendation records, thereby improving the executability of the recommendation task.

[0032] Finally, this application generates a test preparation path recommendation record based on the phased test preparation path sequence and question type matching record, and updates the test preparation profile record based on the execution feedback of the test preparation path recommendation record. This forms a closed-loop processing chain from test preparation profile record, test point priority profile, pre-correction items, phased test preparation path sequence, question type matching record to test preparation path recommendation record, and then the execution feedback writes back to the test preparation profile record. Traditional solutions usually only record the completion status after the recommendation task is issued, making it difficult to promptly reflect the execution feedback on subsequent recommendation criteria. However, this application updates the test preparation profile record with the execution feedback, allowing the updated test preparation profile record to participate in the next test preparation path recommendation cycle. This ensures that the generation of subsequent test point priority profiles can reflect the status changes of the already executed test preparation path recommendation records. Therefore, this application can continuously adjust the test preparation path recommendation record according to the execution feedback of the test preparation object with relatively low operational complexity, improving the continuity and dynamic adaptability of the test preparation path recommendation process. Attached Figure Description

[0033] Figure 1 This application provides an embodiment of a test preparation path recommendation method. Figure 2 This application provides an embodiment of a test preparation path recommendation device. Figure 3 An electronic device is described in an embodiment of this application; Figure 4 This application provides an embodiment of a computer-readable storage medium. Figure 5 This application provides an example of a test preparation path recommendation system. Detailed Implementation

[0034] like Figure 1 The image shows an embodiment of a test preparation path recommendation method. This method is designed for middle school test preparation scenarios and is particularly suitable for scenarios where the target test scope has been determined, the test point dependency graph can express the prior learning relationships between candidate test points, and the test taker continuously generates learning behavior records and answer feedback records during the test preparation process. Figure 1 The process illustrated uses the exam preparation path recommendation cycle as the trigger. It converts learning behavior records, answer feedback records, mastery status records, and historical review time records into exam preparation profile records. It also converts the exam preparation calendar status into the current exam preparation stage record and the target exam configuration into the target exam scope record. Then, it reads the exam point dependency graph based on the target exam scope record. On this basis, the exam preparation profile records participate in the generation of exam point priority profiles, the exam point dependency graph participates in the generation of pre-correction items, the current exam preparation stage record participates in the generation of the stage exam preparation path sequence, and the stage exam preparation path sequence participates in the generation of question type matching records and exam preparation path recommendation records. The intermediate records in this process are not isolated data, but rather continuation data formed by the results of previous processing participating in subsequent processing, allowing the exam preparation path recommendation records to continue to update with the execution feedback of the exam preparation path recommendation records.

[0035] Specifically, in this application, such as Figure 1 As shown, the recommended preparation path includes: When the recommended preparation path cycle is reached, obtain the preparation profile record, current preparation stage record, target exam scope record, and test point dependency graph corresponding to the target exam scope record of the candidate. Based on the test preparation profile records, test point priority status identifiers are set for candidate test points in the test point dependency graph to generate test point priority profiles. Based on the test point dependency graph, a pre-readiness check is performed on the test point priority profile, and pre-correction entries are generated for candidate test points that do not meet the pre-readiness conditions. Based on the current preparation stage records, the priority profiles of test points and the pre-correction items are adapted to the stage to generate a stage preparation path sequence. Based on the sequence of preparation paths, obtain the semantic records of questions, align the semantic records of questions with the semantic records of test points, and generate question type matching records. Based on the sequence of preparation paths and the matching records of question types, a preparation path recommendation record is generated, and the preparation profile record is updated based on the execution feedback of the preparation path recommendation record.

[0036] Specifically, in this application, the exam preparation path recommendation period is used to limit the execution timing of the exam preparation path recommendation method. The exam preparation path recommendation period can be configured according to the exam preparation calendar status or according to the arrival status of the execution feedback of the exam preparation path recommendation records. Taking the exam preparation calendar status configuration as an example, the exam preparation path recommendation period can correspond to the daily exam preparation task generation time, the weekly stage review time, or the first recommendation time after a stage switch; taking the execution feedback arrival status configuration of the exam preparation path recommendation records as an example, the exam preparation path recommendation period can be triggered after the exam preparation object completes the previous round of exam preparation path recommendation records. After the exam preparation path recommendation period arrives, the exam preparation profile record, the current exam preparation stage record, the target exam scope record, and the exam point dependency graph are first obtained. This is to ensure that subsequent exam point selection, exam point priority status flag setting, pre-readiness verification, and stage adaptation all have input basis under the same exam preparation object.

[0037] Specifically, in this application, the test preparation profile record is used to characterize the learning status that a test taker has already formed in the test preparation system. The test preparation profile record is not a single academic result, but rather a data record formed by the structured transformation of learning behavior records, answer feedback records, mastery status records, and historical review time records. Learning behavior records can originate from the test taker's actions in the test preparation system, such as browsing test points, watching explanations, entering exercises, saving questions, and searching for test points; answer feedback records can originate from the test taker's answer results, error annotations, analysis viewing, and correction results during the question-answering process; mastery status records can originate from the test taker's historical answer performance and review performance at the test point level; and historical review time records can originate from the timestamps of the test taker's most recent learning, most recent practice, or most recent review of each test point. After the above records are read when the test preparation path recommendation period arrives, they are not used directly as the path ranking result. Instead, they are first converted into behavior attention markers, error cause pointing markers, mastery gap markers, and review interval markers. This allows the test preparation profile records to be mapped to candidate test points one by one during the subsequent setting of test point priority status identifiers.

[0038] Specifically, in this application, the current preparation stage record is used to characterize the current preparation stage of the candidate and the corresponding task control requirements. The current preparation stage record can be generated from the preparation calendar status, which may include the time position relative to the target exam date, the current review round, the daily task capacity, and stage switching information. The current preparation stage record may include stage task boundaries, exam point coverage strategies, and path length control identifiers. Stage task boundaries are used to limit the range of exam points and task types allowed to enter the stage adaptation in the current preparation stage; the exam point coverage strategy describes whether the current preparation stage leans more towards basic coverage, topical reinforcement, or rapid review; the path length control identifier limits the upper limit of the number of exam points and the capacity of candidate questions in a single preparation path recommendation record. The current preparation stage record is first obtained in the main process of claim 1, and then participates in the formation of the stage preparation path sequence during the stage adaptation process. Therefore, the current preparation stage record is not simply a stage name, but rather a data basis for exam point selection, exam point priority status identifier adjustment, and path length control.

[0039] Specifically, in this application, the target exam scope record is used to limit the exam scope corresponding to the recommended preparation path. The target exam scope record can be determined by the target exam configuration of the candidate, which may include exam subjects, exam grade level, exam version, exam syllabus scope, and the boundaries of this round of preparation tasks. The purpose of the target exam scope record is to limit the reading range of the exam point dependency graph, avoiding the unbounded reading of candidate exam points from all subject exam points. The exam point dependency graph is used to express the prior learning relationships between exam points within the target exam scope record. Candidate exam points in the exam point dependency graph are those that can enter the exam point priority profile processing. The prior exam point chain in the exam point dependency graph represents the prior exam points that candidate exam points depend on in the learning order. The upstream prior exam point chain in the exam point dependency graph represents the more basic exam points that the prior exam points themselves continue to depend on. After reading the test point dependency graph through the target test scope record, the candidate test points, the preceding test point chain, and the upstream preceding test point chain all correspond to the same target test scope record. As a result, the subsequent test point priority profile and preceding correction entries will not deviate from the target test scope record.

[0040] Optionally, the system obtains the candidate's preparation profile record, current preparation stage record, target exam scope record, and test point dependency graph corresponding to the target exam scope record, including: Read the learning behavior records of the test takers from the test preparation system and convert the learning behavior records into behavior attention tags; Read the test-taking feedback records of the test takers during the question-answering process from the test preparation system, and convert the feedback records into error reason markers; Read the mastery status records of test takers in the test point dimension from the test preparation system, and convert the mastery status records into mastery gap markers; Read the historical review time records of the test takers in the test point dimension from the test preparation system, and convert the historical review time records into review interval markers; Combine behavioral attention markers, error cause markers, mastery gap markers, and review interval markers to create a test preparation profile record; The current preparation stage is determined based on the preparation calendar status corresponding to the recommended preparation path cycle. The target exam scope is determined based on the target exam configuration corresponding to the preparation target. The exam point dependency graph is then obtained based on the target exam scope record.

[0041] Specifically, in this application, when converting learning behavior records into behavior attention tags, the learning behavior records can first be merged according to the test point identifier, and then the merged learning behavior records can be encoded by behavior type. Behavior type encoding can include browsing, practice, retrieval, and collection encodings, each of which is bound to a test point identifier. Among these, browsing, practice, retrieval, and collection encodings are all subordinate implementations of behavior type encoding, used to distinguish the source of attention to candidate test points from different operational events in the learning behavior records. After chronologically organizing the behavior type codes corresponding to the same test point identifier, a behavior attention tag corresponding to that test point identifier can be formed. The behavior attention tag is used to indicate the active access and continuous contact status of a test point by the test-taker during a preparation period. After the behavior attention tag is formed, it will be written into the test-taker profile record and mapped to candidate test points in subsequent steps, thus participating in the formation of the test point status mapping record and the test point priority status identifier.

[0042] Specifically, in this application, when converting answer feedback records into error cause pointing tags, the question identifier, test point identifier, answer result, parsing viewing status, and correction status can be read from the answer feedback records first. The question identifier is used to locate the question corresponding to the answer feedback record, the test point identifier is used to locate the test point corresponding to the answer feedback record, the answer result is used to indicate whether the question was passed, the parsing viewing status is used to indicate whether the test taker has opened the question parsing, and the correction status is used to indicate whether the test taker has completed the reprocessing of the incorrect question. After classifying the error causes of answer feedback records under the same test point identifier, error cause pointing tags corresponding to that test point identifier can be formed. Error cause pointing tags do not only record whether the question is wrong, but also convert the answer feedback record into an expression of the error source oriented towards the test point. For example, error causes related to concept recognition, condition extraction, calculation process, and misreading can all be used as the value source of error cause pointing tags; among them, error causes related to concept recognition, condition extraction, calculation process, and misreading all participate in the setting of subsequent test point priority status identifiers through error cause pointing tags. Once the error cause marker is formed, it will be entered into the test preparation profile record along with the behavior attention marker, so that the test preparation profile record can reflect the source of the test point weakness when setting the priority of the candidate test points in the later stage.

[0043] Specifically, in this application, when converting mastery status records into mastery gap markers, the candidate's historical pass status, continuous answer status, corrected answer status, and similar question transfer status can be read at the test point level. Historical pass status indicates the candidate's pass rate on historical questions; continuous answer status indicates the candidate's stable answer rate on consecutive questions; corrected answer status indicates the candidate's answer rate after correction; and similar question transfer status indicates the candidate's transfer answer rate on similar questions. Mastery status records can be stored hierarchically or in intervals. To facilitate subsequent pre-readiness verification, when converting mastery status records into mastery gap markers, the test point identifier, mastery status level, and mastery status update time need to be retained. Mastery gap markers indicate the positions where the candidate still needs to improve at the test point level. After the gap marker is formed, it will participate in two subsequent processes: First, after the gap marker is mapped to the candidate test point, it will participate in the setting of the test point priority status identifier; Second, the gap marker is used to identify the mastery status of each test point in the preceding test point chain, so as to complete the preceding readiness verification in conjunction with the preceding readiness conditions.

[0044] Specifically, in this application, when converting historical review time records into review interval markers, the most recent study time, most recent practice time, and most recent review time for each test point can be read first. The most recent study time, most recent practice time, and most recent review time are all time sources for historical review time records, used to distinguish the time status of the same test point under different review actions. After classifying multiple time records for the same test point by time type, the corresponding historical review time record for that test point can be obtained. The historical review time record is then converted into an interval status based on the current recommended time corresponding to the test preparation path recommendation cycle, forming a review interval marker. The current recommended time is the time base when the test preparation path recommendation cycle arrives; the current recommended time is used to enable the historical review time record to be converted into a review interval marker corresponding to the current test preparation path recommendation cycle. The review interval marker is used to indicate the interval status of the candidate test point from the last effective review. After the review interval marker is formed, it is combined with behavioral attention markers, error cause pointing markers, and mastery gap markers to form a test preparation profile record. Through this combination, the test preparation profile can simultaneously record the test taker's attention status, error sources, mastery gaps, and review intervals, providing multi-faceted data support for setting subsequent test point priority status indicators.

[0045] Specifically, in this application, the exam preparation calendar status is used to determine the current exam preparation stage record. The exam preparation calendar status can consist of the target exam date, the current recommended time, stage boundary configuration, and stage switching records. The stage boundary configuration can be pre-configured according to the exam preparation task requirements, for example, dividing the exam preparation process into a basic coverage stage, a topical reinforcement stage, and a pre-exam review stage; different exam preparation targets can have different stage boundary configurations corresponding to their target exams. When determining the current exam preparation stage record based on the exam preparation calendar status corresponding to the exam preparation path recommendation cycle, the position of the exam preparation path recommendation cycle in the exam preparation calendar status is first read, then that position is mapped to a current exam preparation stage, and subsequently, the stage task boundary, exam point coverage strategy, and path length control identifier corresponding to that current exam preparation stage are read. After the current exam preparation stage record is formed, stage adaptation processing is performed on the exam point priority profile and pre-correction items during stage adaptation, and the stage adaptation processing forms a stage exam preparation path sequence.

[0046] Specifically, in this application, the target exam configuration is used to determine the target exam scope record. The target exam configuration may include the target subject, target exam version, target exam scope code, and the grade level of the test taker. When determining the target exam scope record based on the target exam configuration, the target exam scope code can be read first, then mapped to the test point identifier in the test point directory, and these test point identifiers can be combined into a target exam scope record. After obtaining the target exam scope record, the test point dependency graph can be read based on the target exam scope record. The test point dependency graph can be pre-generated from subject teaching and research data and question bank test point annotation data. The subject teaching and research data is used to provide the prior learning relationships between test points, and the question bank test point annotation data is used to provide the attribution relationships between questions and test points; the subject teaching and research data and the question bank test point annotation data together form the configuration source of the test point dependency graph. Each candidate test point in the test point dependency graph has a test point identifier, test point name, test point level, test point semantic record, and prior test point chain. Once the test point dependency map is read, it will participate in subsequent processing together with the test preparation profile record and the current test preparation stage record, so that the priority status and pre-correction of the candidate test points are limited to the target test range record.

[0047] Specifically, in this application, Figure 1 The test preparation profile records, current test preparation stage records, target exam scope records, and test point dependency graphs can be linked using a unified data identifier. This data identifier can be formed by combining a test preparation object identifier, a target exam configuration identifier, and a test preparation path recommendation cycle identifier. The test preparation object identifier is used to locate the test preparation object, the target exam configuration identifier is used to locate the target exam configuration, and the test preparation path recommendation cycle identifier is used to locate the current test preparation path recommendation cycle. The data identifier is used to locate the test preparation profile records, current test preparation stage records, target exam scope records, and test point dependency graphs of the same test preparation object within the same test preparation path recommendation cycle. Through this data identifier, the subsequently generated test point priority profiles, pre-correction entries, and stage test preparation path sequences can all be traced back to the same input source, reducing data mixing between different test preparation cycles or different target exam scopes; at the same time, this data identifier will continue to be written into the test point priority profiles, pre-correction entries, and stage test preparation path sequences to ensure that the test point priority profiles, pre-correction entries, and stage test preparation path sequences are consistent with their source data.

[0048] Optionally, based on the test preparation profile records, test point priority status identifiers are set for candidate test points in the test point dependency graph, generating a test point priority profile, including: In the test point dependency graph, candidate test points are determined based on the target test scope record; The behavioral attention markers, error cause markers, mastery gap markers, and review interval markers in the test preparation profile are mapped to candidate test points to obtain test point status mapping records corresponding to the candidate test points. Based on the test site status mapping record, generate the test site priority status identifier corresponding to the candidate test site, and generate the test site priority profile based on the candidate test site carrying the test site priority status identifier.

[0049] Specifically, in this application, candidate test points are the test points to be recommended located in the test point dependency graph of the target exam scope record. When determining candidate test points, test point identifiers within the target exam scope can be read first from the target exam scope record, and then the test point nodes corresponding to these identifiers can be queried in the test point dependency graph. The queried test point nodes are used as candidate test points. Each candidate test point has a test point identifier, test point name, test point level, test point semantic record, and preceding test point chain. Candidate test points are not immediately included in the recommended study path record, but rather are initially included in the test point priority profile. Using candidate test points as pending test points ensures that subsequent behavior attention markers, error cause pointing markers, mastery gap markers, and review interval markers are all mapped around the candidate test points, avoiding invalid processing of test points outside the target exam scope record. After mapping, candidate test points form a test point status mapping record, which is then used to generate test point priority status identifiers.

[0050] Specifically, in this application, when mapping behavioral attention markers to candidate test points, the test point identifier in the behavioral attention marker can be matched with the test point identifier of the candidate test point. If the test point identifier in the behavioral attention marker matches the test point identifier of the candidate test point, the behavioral attention marker is written into the test point status mapping record corresponding to the candidate test point; if the test point identifier in the behavioral attention marker corresponds to the preceding test point chain of the candidate test point, the behavioral attention marker is written into the preceding attention field corresponding to the candidate test point to reflect the candidate's access to the basic content of the candidate test point. When mapping error cause pointing markers to candidate test points, matching can be performed according to the same test point identifier or according to the test point affiliation relationship between the question semantic record and the test point semantic record; the matched error cause pointing marker is written into the error cause pointing field in the test point status mapping record. When mapping mastery gap markers to candidate test points, the mastery status level and mastery status update time need to be retained for subsequent preceding readiness verification; the matched mastery gap marker is written into the mastery gap field in the test point status mapping record. When mapping review interval markers to candidate test points, the test point identifier and interval status need to be retained. The interval status reflects whether the candidate test point needs to be revisited within the current preparation path recommendation period. The matched review interval markers are written to the review interval field in the test point status mapping record. After the above mapping process is completed, the test point status mapping record corresponding to the candidate test point will include the combined status of behavioral attention markers, error cause pointing markers, mastery gap markers, and review interval markers under the same candidate test point.

[0051] Specifically, in this application, the test point status mapping record is used to convert different markers in the test preparation profile record into a unified status expression on candidate test points. The test point status mapping record may include candidate test point identifiers, behavioral focus fields, error cause pointing fields, mastery gap fields, and review interval fields. The behavioral focus field, derived from the behavioral focus marker, indicates the intensity of the test preparation candidate's contact with the candidate test point; the error cause pointing field, derived from the error cause pointing marker, indicates the main source of errors related to the candidate test point; the mastery gap field, derived from the mastery gap marker, indicates the mastery status of the candidate test point; and the review interval field, derived from the review interval marker, indicates the interval between the candidate test point and the last effective review. After the test point status mapping record is formed, it is not directly used as a recommendation result, but rather used to set the test point priority status identifier; after the test point priority status identifier is formed, it will continue to be written into the test point priority profile along with the candidate test points and the test point status mapping record.

[0052] Specifically, in this application, the test site priority status identifier can be generated through pre-configured priority status generation rules. These rules can include marker source identification rules, marker conflict handling rules, and status synthesis rules. The marker source identification rules identify which type of marker in the test preparation profile record the behavioral attention field, error cause pointing field, mastery gap field, and review interval field originates from. The marker conflict handling rules handle situations where different markers for the same candidate test site indicate inconsistent priority directions. For example, if the behavioral attention field shows that the test preparation candidate has frequently visited the candidate test site, while the mastery gap field still indicates a mastery gap exists, the mastery gap field can be given a higher processing position in status synthesis. The status synthesis rules combine the behavioral attention field, error cause pointing field, mastery gap field, and review interval field into a test site priority status identifier. The test site priority status identifier can be stored as a level identifier, sorting identifier, or task status identifier, with the specific format chosen based on the data structure of the test preparation system. In this way, the test site priority status identifier is not a single score but rather a stateful result that retains multiple types of profile marker sources. Priority status generation rules can be pre-stored in the exam preparation system. The configuration of priority status generation rules includes the reading order of each field, the conflict handling order of each field, and the status output format when each field is combined into the exam point priority status identifier. When the exam preparation path recommendation cycle arrives, the priority status generation rules are read and applied to the exam point status mapping record.

[0053] Specifically, this application can also use a multi-dimensional priority scoring system as an example of an implementation for identifying the priority status of test points. This multi-dimensional priority scoring system can be formed by learning behavior weights, knowledge point mastery weights, time decay weights, and personalized adjustment weights. Learning behavior weights are derived from behavior attention markers, knowledge point mastery weights are derived from mastery gap markers and error cause pointing markers, time decay weights are derived from review interval markers, and personalized adjustment weights can be derived from the target subject bias or weak module configuration in the target exam configuration. To facilitate implementation in a computer, the above weights can first be converted into weight states within the same value range, and then synthesized according to the corresponding weight ratios recorded in the current preparation stage. Taking high school mathematics preparation as an example, if a candidate test point is a function extremum, the behavior attention marker indicates that the test taker has recently entered the analysis page of this candidate test point multiple times, the error cause pointing marker indicates that there are conceptual comprehension errors under this candidate test point, the mastery gap marker indicates that this candidate test point has not yet met the requirements of the current stage, and the review interval marker indicates that the candidate test point has crossed the current review cycle since the last review. In this case, the priority status generation rule will cause this candidate test point to form a higher priority status identifier. The test site priority status identifier is then written into the test site priority profile for retrieval during pre-test readiness verification. The multi-dimensional priority scoring method is used in this application to illustrate one method of generating the test site priority status identifier, which still participates in the subsequent test site priority profile and pre-test readiness verification under the name stated in the claims.

[0054] Specifically, in this application, the test point priority profile is a structured record composed of multiple candidate test points carrying test point priority status identifiers. The test point priority profile includes at least candidate test points, test point status mapping records, and test point priority status identifiers. After the test point priority profile is formed, it enters the pre-readiness verification stage. Because the test point priority profile retains the test point identifiers of the candidate test points, the pre-readiness verification can read the corresponding pre-test point chain from the test point dependency graph based on these identifiers; because the test point priority profile retains the test point priority status identifiers, the pre-readiness verification can update the path position of the candidate test point when it is downgraded or strengthened; because the test point priority profile retains the test point status mapping records, the pre-readiness verification can identify the mastery status in the pre-test point chain based on the mastery gap markers in the test preparation profile record. Therefore, the test point priority profile continues to participate in subsequent pre-readiness verifications after its generation, rather than remaining at the level of ranking results.

[0055] Optionally, based on the test point dependency graph, a pre-readiness check is performed on the test point priority profile, and pre-correction entries are generated for candidate test points that do not meet the pre-readiness conditions, including: Read the candidate test points carrying the test point priority status identifier from the test point priority profile, and obtain the preceding test point chain corresponding to the candidate test point from the test point dependency graph; Based on the test preparation profile record, determine the mastery status of each pre-test point in the pre-test point chain, and perform pre-test point readiness verification on the pre-test point chain according to the preset pre-test readiness conditions. If there is a candidate test point in the current test point chain that does not meet the prerequisite readiness conditions, set a downgrade flag for the candidate test point and identify the candidate test point that does not meet the prerequisite readiness conditions as a prerequisite reinforcement item. The candidate test point is generated based on a combination of the downgrade identifier and the preceding reinforcement items.

[0056] Specifically, in this application, the pre-readiness check is used to determine whether the prerequisite test points on which a candidate test point depends have the basic state to enter the subsequent learning tasks before the candidate test point enters the preparation path sequence. The inputs to the pre-readiness check include a test point priority profile, a test point dependency graph, and a preparation profile record. The test point priority profile provides candidate test points carrying test point priority status identifiers, the test point dependency graph provides the prerequisite test point chain corresponding to the candidate test point, and the preparation profile record provides the mastery status of each prerequisite test point in the prerequisite test point chain. Through the cooperation of these three inputs, the pre-readiness check can combine the dependency relationship between test points with the current mastery status of the preparation object, avoiding path generation based solely on the priority status of the candidate test point itself. After the pre-readiness check is completed, the pre-readiness check result is obtained, which is used to determine whether a candidate test point generates a pre-correction entry.

[0057] Specifically, in this application, the prerequisite test point chain can be formed according to the directed succession relationship in the test point dependency graph. Each prerequisite test point in the prerequisite test point chain has a prerequisite test point identifier and a prerequisite relationship type. The prerequisite relationship type can represent a necessary foundation, direct succession, or same-level support relationship; necessary foundation, direct succession, and same-level support are all subordinate implementations of the prerequisite relationship type, which is used to describe the succession method between the prerequisite test point and the candidate test point in the prerequisite test point chain. After the prerequisite test point chain is read, the prerequisite test point identifier needs to be matched with the mastery gap mark in the test preparation profile record. If the prerequisite test point identifier can be matched with the mastery gap mark in the test preparation profile record, the mastery status of the prerequisite test point is determined according to the mastery gap mark; if the prerequisite test point identifier does not match the mastery gap mark, the historical mastery status in the test preparation profile record can be read as the mastery status of the prerequisite test point. The historical mastery status is derived from the existing mastery status records of the preceding test point in the test preparation profile. After being read, the historical mastery status is used to supplement the mastery status of the preceding test points in the preceding test point chain. In this way, each preceding test point in the preceding test point chain can obtain a mastery status related to the test taker, and the mastery status then participates in the judgment of the preceding ready conditions.

[0058] Specifically, in this application, prerequisite readiness conditions can be pre-configured in the test point dependency graph or the current preparation stage record. Prerequisite readiness conditions are used to determine whether the prerequisite test points in the prerequisite test point chain meet the basic requirements for entering the candidate test point learning stage. Prerequisite readiness conditions may include mastery level conditions, recent answer pass conditions, and review interval conditions. The mastery level condition is used to determine whether the mastery status of the prerequisite test point has reached the target level; the recent answer pass condition is used to determine whether the prerequisite test point has a continuous failure status in recent question answers; the review interval condition is used to determine whether the prerequisite test point needs to be prioritized for revisiting due to a long period of inactivity. The mastery level condition, recent answer pass condition, and review interval condition are all subordinate implementations of the prerequisite readiness conditions, and the prerequisite readiness conditions use these subordinate implementations to uniformly determine the prerequisite test points in the prerequisite test point chain. Prerequisite readiness conditions can be configured differently in different current preparation stage records; for example, a broader set of prerequisite readiness conditions can be used in the basic coverage stage, while a stricter set of prerequisite readiness conditions can be used in the topic reinforcement stage. After the prerequisite readiness conditions are read, they will be applied to each prerequisite test point in the prerequisite test point chain to obtain the prerequisite readiness verification result; the prerequisite readiness verification result will continue to participate in the generation of the downgrade flag and the prerequisite reinforcement entry.

[0059] Specifically, in this application, when there are candidate test points in the current test point chain that do not meet the prerequisite readiness conditions, the corresponding candidate test point is not directly deleted. Instead, a downgrade flag is set for the candidate test point, and the candidate test point that does not meet the prerequisite readiness conditions is identified as a prerequisite reinforcement entry. The downgrade flag is used to change the arrangement position or entry condition of the corresponding candidate test point in the stage preparation path sequence, and the prerequisite reinforcement entry is used to retain the candidate test point in the form of an insertable path. The downgrade flag and the prerequisite reinforcement entry respectively express two types of correction actions for the same candidate test point: the downgrade flag acts on the candidate test point itself, and the prerequisite reinforcement entry acts on the prerequisite test points that the candidate test point depends on. The combination of the two forms the prerequisite correction entry corresponding to the candidate test point, so that the candidate test point can reflect both its own priority status and the path correction relationship caused by its insufficient prerequisite foundation during subsequent stage adaptation.

[0060] Specifically, in this application, the pre-correction entries may include at least candidate test point identifiers, downgrade identifiers, pre-correction reinforcement entries, and verification source identifiers. The candidate test point identifier points to candidate test points in the test point priority profile; the downgrade identifier indicates that the candidate test point needs to adjust its path position due to failing the pre-correction readiness verification; the pre-correction reinforcement entry records the pre-correction test points that do not meet the pre-correction readiness conditions; and the verification source identifier records which pre-correction test point chain the pre-correction entry originates from. After the pre-correction entries are formed, they enter the stage adaptation phase together with the test point priority profile. During stage adaptation, when generating the stage preparation path sequence, the pre-correction reinforcement entries are placed before the corresponding candidate test points, and the downgrade identifier is applied to the corresponding candidate test points. Thus, the pre-correction entries continue to participate in subsequent path processing after their formation, and the results obtained from the pre-correction readiness verification can be transformed into order constraints in the stage preparation path sequence.

[0061] Specifically, this application uses high school mathematics functions and derivatives exam preparation as an example to illustrate the formation process of prerequisite correction entries. If a candidate exam point in the exam point priority profile is the proof of derivative application inequalities, the exam point dependency graph reads the prerequisite exam point chain corresponding to this candidate exam point, which includes basic derivative operations, function monotonicity, and basic inequality properties. If the mastery gap marker in the exam preparation profile record shows that function monotonicity does not meet the prerequisite readiness condition, then function monotonicity is identified as a prerequisite reinforcement entry, and a downgrade marker is set for the proof of derivative application inequalities. Subsequently, the downgrade marker and the prerequisite reinforcement entry are combined to generate the prerequisite correction entry corresponding to the proof of derivative application inequalities. After this prerequisite correction entry enters the stage adaptation, function monotonicity will be arranged before the proof of derivative application inequalities in the stage exam preparation path sequence, and the proof of derivative application inequalities will adjust its path position according to the downgrade marker. Through this process, the knowledge connection relationship between exam points is transformed into an executable path sequence relationship.

[0062] Optionally, prerequisite reinforcement items participate in the generation of the stage preparation path sequence, including: After generating the prerequisite reinforcement items, determine whether there are upstream prerequisite test points in the prerequisite reinforcement items; If it exists, the corresponding upstream prerequisite test point chain is obtained according to the test point dependency graph, and the mastery status of each upstream prerequisite test point in the upstream prerequisite test point chain is determined according to the test preparation profile record. Based on the prerequisite readiness conditions, perform prerequisite readiness checks on the upstream prerequisite test point chain, and update the prerequisite reinforcement entries according to the check results. Set the pre-readiness verification completed pre-compliance entries to an insertable state, so that pre-compliance entries carrying the insertable state are arranged before the corresponding candidate test points during stage adaptation.

[0063] Specifically, in this application, after the pre-examination reinforcement items are formed, it is also necessary to determine whether there are upstream pre-examination points in the pre-examination reinforcement items. An upstream pre-examination point refers to a basic examination point that the pre-examination point in the pre-examination reinforcement item itself continues to depend on. This determination can be made through the directed succession relationship in the examination point dependency graph. If there is an upstream edge pointing to the pre-examination point in the examination point dependency graph, then the examination point corresponding to that upstream edge constitutes an upstream pre-examination point; if there are multiple upstream pre-examination points, then these upstream pre-examination points form an upstream pre-examination point chain according to the succession order in the examination point dependency graph. After the upstream pre-examination point chain is read, it will continue to be matched with the test preparation profile record to determine the mastery status of each upstream pre-examination point in the upstream pre-examination point chain. The mastery status of each upstream pre-examination point in the upstream pre-examination point chain continues to participate in the judgment of the pre-examination readiness conditions and forms the upstream verification result.

[0064] Specifically, in this application, the pre-readiness verification of the upstream pre-examination point chain and the pre-readiness verification of the pre-examination point chain use the same pre-readiness condition. This is because before a pre-examination reinforcement entry is inserted into the stage preparation path sequence, the pre-examination reinforcement entry itself must also satisfy the basic continuity relationship; otherwise, the stage preparation path sequence will still have path positions with insufficient foundation. After performing the pre-readiness verification on the upstream pre-examination point chain, the upstream verification result can be obtained. If the upstream verification result shows that all upstream pre-examination points in the upstream pre-examination point chain satisfy the pre-readiness condition, the pre-examination reinforcement entry retains its original content, and the upstream verification result is written into the pre-examination reinforcement entry; if the upstream verification result shows that there are upstream pre-examination points that do not satisfy the pre-readiness condition, the upstream pre-examination points that do not satisfy the pre-readiness condition are written into the pre-examination reinforcement entry, expanding the pre-examination reinforcement entry from the original pre-examination point to a reinforcement entry containing upstream pre-examination points. The updated pre-examination reinforcement entry will then enter the insertable state judgment.

[0065] Specifically, in this application, the insertable state indicates that a pre-reinforcement entry has completed the pre-readiness verification of its upstream pre-examination point chain and can participate in the generation of the stage preparation path sequence. When setting a pre-reinforcement entry that has completed the pre-readiness verification to the insertable state, the insertable state needs to be written into the pre-reinforcement entry, and an insertion relationship needs to be established between the pre-reinforcement entry carrying the insertable state and the corresponding candidate examination point. This insertion relationship can include the corresponding candidate examination point identifier, the insertion position identifier, and the reinforcement order identifier. The corresponding candidate examination point identifier is used to point to the candidate examination point that needs to be reinforced; the insertion position identifier is used to indicate that the pre-reinforcement entry should be located before the corresponding candidate examination point during stage adaptation; the reinforcement order identifier is used to indicate the arrangement order among multiple pre-examination points in the pre-reinforcement entry. Through the insertable state and the insertion relationship, the pre-reinforcement entry can be transformed into a pre-task in the stage preparation path sequence during stage adaptation, instead of remaining in the verification result.

[0066] Specifically, in this application, continuing with the aforementioned high school mathematics example, if the prerequisite reinforcement item is the monotonicity of a function, and the test point dependency graph shows that the function monotonicity has an upstream prerequisite test point—the basic meaning of the derivative—then the mastery status of the basic meaning of the derivative is determined based on the test preparation profile record. If the basic meaning of the derivative does not meet the prerequisite readiness conditions, then the basic meaning of the derivative is added to the prerequisite reinforcement item, and arranged before the function monotonicity according to the succession relationship in the test point dependency graph. After the prerequisite readiness verification of the basic meaning of the derivative and the function monotonicity in the prerequisite reinforcement item is completed, the prerequisite reinforcement item is set to an insertable state. During subsequent stage adaptation, the prerequisite reinforcement item carrying the insertable state will be arranged before the proof of the derivative application inequality, thus making the stage test preparation path sequence present an order of first strengthening the basics and then entering the target candidate test points.

[0067] Specifically, in this application, through the processing relationship between pre-reinforcement entries, upstream pre-examination point chains, and insertable states, pre-correction entries can express multi-layered pre-reinforcement relationships. Pre-reinforcement entries are used to carry pre-examination points that do not meet the pre-readiness conditions; upstream pre-examination point chains are used to further verify whether the pre-examination points in the pre-reinforcement entries have more basic examination point conditions; and insertable states are used to indicate whether pre-reinforcement entries can participate in the generation of the stage preparation path sequence. The above processing enables pre-reinforcement entries to inherit necessary basic examination points layer by layer from the examination point dependency graph and transform the inheritance result into the path order in the stage preparation path sequence. Compared with the method of simply performing a descending order processing on candidate examination points, this application can transform the pre-reinforcement relationship between examination points into insertable path entries before generating the stage preparation path sequence, making the examination point order in the preparation path recommendation record more consistent with the knowledge inheritance relationship within the target examination scope record.

[0068] Optionally, based on the records of the current preparation stage, the priority profile of test points and the pre-exam correction items are adapted to the current stage to generate a sequence of preparation paths for each stage, including: Extract the phase task boundaries, test point coverage strategies, and path length control identifiers from the current preparation phase records; Based on the phase task boundaries, candidate test points for each phase are selected from the test point priority profile; The test site priority status identifier is obtained by adjusting the test site priority status identifier of the candidate test sites in the stage according to the test site coverage strategy. Based on the path length control identifier, a portion of the candidate test points carrying the stage priority status identifier are extracted, and a stage preparation path sequence is generated according to the order in which the pre-correction reinforcement entries in the pre-correction entries are located before the corresponding candidate test points, and the demotion identifiers in the pre-correction entries are applied to the corresponding candidate test points.

[0069] Specifically, in this application, stage adaptation is used to convert the test point priority profile and prerequisite correction items into a stage preparation path sequence that matches the current preparation stage record. The test point priority profile already records candidate test points and their test point priority status identifiers, the prerequisite correction items already record downgrade identifiers and prerequisite reinforcement items, and the current preparation stage record provides stage task boundaries, test point coverage strategies, and path length control identifiers. Stage adaptation does not regenerate candidate test points, nor does it simply sort them according to the test point priority status identifiers. Instead, it applies the stage task boundaries, test point coverage strategies, and path length control identifiers from the current preparation stage record to the test point priority profile and prerequisite correction items, so that the stage preparation path sequence simultaneously reflects the current status of the preparation object, the prerequisite relationships between test points, and the limitations of the current preparation stage record on the path length control identifiers and test point coverage strategies.

[0070] Specifically, in this application, the stage task boundary is used to limit the range of test points allowed to enter the stage preparation path sequence in the current preparation stage record. The stage task boundary can be read from the current preparation stage record, which can be determined by the preparation calendar status. The stage task boundary can include basic test point boundaries, topic test point boundaries, error-prone test point boundaries, and high-frequency test point boundaries; the basic test point boundaries, topic test point boundaries, error-prone test point boundaries, and high-frequency test point boundaries are all subordinate implementations of the stage task boundary, used to limit the entry range of different candidate test points under different current preparation stage records. When filtering stage candidate test points from the test point priority profile according to the stage task boundary, the candidate test points in the test point priority profile are read first, then the test point level, test point category, and test point priority status identifier of the candidate test points are read, and then the candidate test points are matched with the stage task boundary. Candidate test points that match the stage task boundary are written into the stage candidate test points, and candidate test points that do not match the stage task boundary are retained in the test point priority profile for continued use in the next preparation path recommendation cycle. Through this process, the candidate test points for each stage are derived from the test point priority profile, and these candidate test points then continue to participate in the adjustment of the test point coverage strategy.

[0071] Specifically, in this application, the test point coverage strategy is used to describe the coverage direction of the current preparation stage records for stage candidate test points. The test point coverage strategy can include a coverage priority strategy, a weakness-remediation priority strategy, and a revisit priority strategy. The coverage priority strategy ensures a more balanced entry opportunity for stage candidate test points in the basic coverage stage; the weakness-remediation priority strategy prioritizes stage candidate test points corresponding to mastery gap markers, giving them a higher stage priority status; and the revisit priority strategy prioritizes stage candidate test points corresponding to review interval markers, placing them in the review position. After reading from the current preparation stage records, the test point coverage strategy adjusts the test point priority status of stage candidate test points, and the adjusted test point priority status is written into the stage priority status. The stage priority status remains bound to the corresponding stage candidate test point and is used as the sorting basis when the subsequent path length control marker truncates stage candidate test points. Through the participation of the test point coverage strategy, stage candidate test points are not only selected from the test point priority profile but are also transformed into stage candidate test points suitable for the current preparation stage records.

[0072] Specifically, in this application, the path length control identifier is used to limit the number and arrangement length of candidate test points in the stage preparation path sequence. The path length control identifier can originate from the current preparation stage record or be converted from the task capacity record corresponding to the preparation path recommendation cycle. The task capacity record can be pre-configured based on the number of test points allowed to be displayed and the number of candidate questions allowed to be configured within the preparation path recommendation cycle. After conversion, the task capacity record forms the path length control identifier, which participates in the selection of stage candidate test points. After being read during stage adaptation, the path length control identifier is first applied to the stage candidate test points carrying the stage priority status identifier, and then a portion of the test points are selected according to the arrangement order corresponding to the stage priority status identifier. This selection is not arbitrary; rather, it is performed after the stage priority status identifier has been adjusted by the test point coverage strategy, generating the main test point entries of the stage preparation path sequence according to the path length limited by the path length control identifier. The selected stage candidate test points will continue to be combined with the preceding correction entries, while the unselected stage candidate test points are retained in the test point priority profile for re-participation in stage adaptation in subsequent preparation path recommendation cycles.

[0073] Specifically, in this application, the role of preceding correction entries in stage adaptation includes the insertion of preceding reinforcement entries and the function of degradation markers. Preceding a candidate test point means that after stage adaptation reads the preceding correction entries, it arranges the preceding test points in the preceding reinforcement entries before the corresponding candidate test points according to the insertion relationship between the preceding reinforcement entries and the corresponding candidate test points. Degradation markers act on corresponding candidate test points, meaning that after stage adaptation reads the preceding correction entries, it adjusts the entry position or entry state of the corresponding candidate test point in the stage preparation path sequence according to the degradation markers. Both preceding reinforcement entries and degradation markers originate from preceding correction entries, which in turn originate from the pre-readiness verification of the test point priority profile based on the test point dependency graph. Therefore, the order in the stage preparation path sequence is not a simple sorting result, but rather formed by the combined effect of the current preparation stage record, the test point priority profile, and the preceding correction entries.

[0074] Specifically, in this application, the stage preparation path sequence can be stored as path node entries. Each path node entry can include a test point identifier, a stage priority status identifier, a path position identifier, a preceding source identifier, and a degradation effect identifier. The test point identifier points to a stage candidate test point or a preceding test point in a preceding reinforcement entry; the stage priority status identifier indicates the priority status of the stage candidate test point recorded in the current preparation stage; the path position identifier indicates the position of the test point in the stage preparation path sequence; the preceding source identifier indicates whether the test point comes from a preceding reinforcement entry; and the degradation effect identifier indicates whether the test point is affected by the degradation identifier. After the stage preparation path sequence is formed, it will serve as the entry point for subsequently obtaining question semantic records. In other words, the test point identifiers in the stage preparation path sequence will directly participate in the semantic alignment between question semantic records and test point semantic records.

[0075] Specifically, in this application, the high school mathematics function and derivative exam preparation can be used as an example to illustrate the formation process of stage adaptation. If the current exam preparation stage record corresponds to the topic reinforcement stage, the stage task boundary limits the function and derivative-related exam points to enter the stage adaptation, the exam point coverage strategy requires priority processing of exam points corresponding to the gap markers, and the path length control identifier limits the number of exam points carried by the current stage preparation path sequence. If the stage candidate exam points in the exam point priority profile include function extrema, function monotonicity, and derivative application inequality proofs, and the pre-correction entry shows that function monotonicity is a pre-reinforcement entry for derivative application inequality proofs, and derivative application inequality proofs carry a downgrade identifier, then the stage adaptation will first retain the above stage candidate exam points according to the stage task boundary, then adjust the exam point priority status identifier of the above stage candidate exam points according to the exam point coverage strategy to obtain the stage priority status identifier, and then extract the stage candidate exam points carrying the stage priority status identifier according to the path length control identifier, and arrange function monotonicity before derivative application inequality proofs to generate the stage preparation path sequence. Therefore, the phased preparation path sequence can both respond to the current preparation phase record and retain the path order constraints formed by the previous correction items.

[0076] Specifically, in this application, after phase adaptation is completed, the phase preparation path sequence serves as the entry point for generating subsequent question type matching records. Each test point identifier in the phase preparation path sequence is used to read the corresponding test point semantic record. The test point semantic record is then used to retrieve candidate questions from the question bank index and align its semantic anchor with the question semantic record of the candidate question. Thus, the phase preparation path sequence is not only the source of the test point order in the preparation path recommendation record but also the entry point for question matching processing in the question type matching record. Through this connection, the processing result of phase adaptation can continue to influence the reading of question semantic records and the generation of preparation path recommendation records.

[0077] Optionally, the current preparation stage record is generated based on the preparation calendar status; when the preparation calendar status indicates a stage switch, the stage task boundary, test point coverage strategy and path length control identifier in the current preparation stage record are updated according to the switched preparation stage. When the next preparation path recommendation cycle arrives, the test point priority profile and pre-correction items will be re-adapted to the current preparation stage based on the updated current preparation stage record, so that the newly generated stage preparation path sequence matches the preparation stage after the switch.

[0078] Specifically, in this application, the exam preparation calendar status is used to describe the position of the exam preparation path recommendation cycle on the exam preparation timeline. The exam preparation calendar status can include the target exam date, the current recommended time, the stage boundary configuration, and the stage switch record. The target exam date provides a baseline for the end of exam preparation, the current recommended time represents the baseline time when the exam preparation path recommendation cycle arrives, the stage boundary configuration defines the time range of different exam preparation stages, and the stage switch record records whether the current exam preparation stage record has been updated. Through the target exam date, the current recommended time, the stage boundary configuration, and the stage switch record, the current exam preparation stage record can be determined, and the stage task boundaries, test point coverage strategies, and path length control identifiers in the current exam preparation stage record can be configured as stage control data suitable for the current recommended time.

[0079] Specifically, in this application, when the exam preparation calendar status indicates a stage switch, the current exam preparation stage record needs to be updated according to the switched exam preparation stage. Stage switching can be triggered by the current recommended time crossing stage boundaries or by execution feedback from the exam preparation path recommendation record. After a stage switch occurs, the stage task boundaries, test point coverage strategies, and path length control identifiers in the current exam preparation stage record all need to be reread. Updating the stage task boundaries changes the range of candidate test points that can enter stage adaptation within the next exam preparation path recommendation cycle; updating the test point coverage strategy changes the way the test point priority status identifiers of candidate test points are adjusted; updating the path length control identifier changes the number of test points and the capacity of candidate questions in the stage exam preparation path sequence. Therefore, updating the current exam preparation stage record directly affects the stage adaptation processing in the next exam preparation path recommendation cycle.

[0080] Specifically, in this application, the switched preparation stages can correspond to different path generation requirements. The basic coverage stage can focus more on broader coverage of basic test points within the target exam scope record; the topic reinforcement stage can focus more on the pre-positioning of paths corresponding to test points marked with gaps and test points marked with error causes; and the pre-exam review stage can focus more on short-path review of test points corresponding to review interval marks and high-frequency test points. The above stage names are only one feasible expression of the current preparation stage record. In specific implementation, the current preparation stage record can also adopt other stage division methods. Regardless of the stage division method adopted, the current preparation stage record participates in stage adaptation through stage task boundaries, test point coverage strategies, and path length control identifiers, and ensures that the stage preparation path sequence matches the switched preparation stage.

[0081] Specifically, in this application, when the next preparation path recommendation cycle arrives, the test point priority profile and preliminary correction items are re-adapted to the current preparation stage record based on the updated record. During this re-adaptation, the test point priority profile still provides candidate test points and test point priority status identifiers, the preliminary correction items still provide preliminary reinforcement items and downgrade identifiers, and the updated current preparation stage record provides new stage task boundaries, test point coverage strategies, and path length control identifiers. The new stage task boundaries are used to re-select stage candidate test points, the new test point coverage strategy is used to readjust the test point priority status identifiers of stage candidate test points, and the new path length control identifiers are used to re-truncate stage candidate test points carrying stage priority status identifiers. The newly generated stage preparation path sequence will match the switched preparation stage and continue to participate in the semantic alignment between the question semantic record and the test point semantic record.

[0082] Specifically, in this application, prior correction entries are not directly discarded before and after stage switching. If prior correction entries have already been generated for candidate test points before stage switching, they will still enter stage adaptation along with the test point priority profile when the next preparation path recommendation cycle arrives. If the requirements for prior readiness conditions change after the switch, prior readiness verification can be re-executed after the current preparation stage record is updated, and the prior correction entries can be updated based on the re-obtained prior readiness verification result. This approach avoids the prior reinforcement entries from becoming disconnected from the current preparation stage record due to stage switching, and also allows the newly generated stage preparation path sequence to continue to retain the prior learning relationships between candidate test points.

[0083] Specifically, this application uses the example of a test-taker switching from the basic coverage stage to the topic-based reinforcement stage to illustrate the updating of the current test-taker's record. In the basic coverage stage, the stage task boundary allows basic and intermediate-difficulty test points from the target exam scope record to enter the stage adaptation. The test point coverage strategy prioritizes uncovered basic test points into the stage test-taker's ...

[0084] Specifically, in this application, after the current preparation stage record is updated, a stage switch source identifier can also be recorded. The stage switch source identifier can indicate that the stage switch is triggered by the preparation calendar state or by the execution feedback of the preparation path recommendation record. After the stage switch source identifier is written into the current preparation stage record, it will participate in the next preparation path recommendation cycle along with the current preparation stage record. If the stage switch source identifier comes from the execution feedback of the preparation path recommendation record, it means that the preparation object has formed a new preparation profile record after the execution of the previous preparation path recommendation record. The next preparation path recommendation cycle will re-form the stage preparation path sequence based on the updated preparation profile record and the updated current preparation stage record. This processing creates a continuous technical connection between the execution feedback of the preparation path recommendation record, the preparation profile record, and the current preparation stage record.

[0085] Optionally, based on the stage preparation path sequence, question semantic records are obtained, and the question semantic records are semantically aligned with the test point semantic records to generate question type matching records, including: Obtain the corresponding semantic records of test points based on the test point identifiers in the phased preparation path sequence; Retrieve candidate questions that match the test point semantic records from the preset question bank index, and extract the question semantic records corresponding to the candidate questions; Semantic anchor point alignment is performed between the semantic records of the questions and the semantic records of the test points to obtain the question matching identifier; Based on the candidate questions carrying question matching identifiers, a question type matching record for the corresponding test point is generated.

[0086] Specifically, in this application, the question type matching record is used to establish a semantic correspondence between test points in the stage preparation path sequence and candidate questions that can be used for training or review. The stage preparation path sequence provides test point identifiers, which are used to read the test point semantic record; the test point semantic record is used to retrieve candidate questions from a pre-defined question bank index; the candidate questions are used to extract question semantic records; the question semantic records are then aligned with the test point semantic records using semantic anchors to form a question matching identifier. After the question matching identifier is formed, it is written to the candidate questions, and the candidate questions carrying the question matching identifier are then written to the question type matching record. Through this process, the question type matching record is not generated solely based on question type tags, but is formed by the combined action of the stage preparation path sequence, the test point semantic record, and the question semantic record.

[0087] Specifically, in this application, the test point semantic records can be pre-configured in the test point dependency graph, or they can be read from the test point semantic database based on the test point identifiers in the test point dependency graph. Test point semantic records can include test point definition fragments, test point condition fragments, test point operation fragments, related concept fragments, and common error cause fragments. Test point definition fragments describe the conceptual boundaries of the test point; test point condition fragments describe the conditional expressions that frequently appear in questions; test point operation fragments describe the problem-solving methods that usually correspond to the test point; related concept fragments record the adjacent test points in the test point dependency graph; and common error cause fragments record the semantic source of the error cause pointing markers corresponding to the test point. After the test point identifiers in the stage preparation path sequence read the test point semantic records, the test point semantic records will serve as the basis for candidate question recall and semantic anchor alignment.

[0088] Specifically, in this application, a pre-defined question bank index is used to retrieve candidate questions from the question bank according to the semantic records of test points. The pre-defined question bank index can be pre-built based on question identifiers, test point annotations, question stem text, option text, explanation text, and question type tags. The configuration process of the pre-defined question bank index can first bind the question identifiers in the question bank with the test point annotations, then store the question stem text, option text, and explanation text in segments, and finally establish an index relationship between the question type tags and the question identifiers. When retrieving candidate questions based on the semantic records of test points, an initial search can be performed using the test point identifiers, followed by a review based on the test point definition fragments, test point condition fragments, and related concept fragments in the test point semantic records to obtain candidate questions that match the test point semantic records. After the candidate questions are read, they do not directly enter the recommended study path records, but instead, the question semantic records are extracted first.

[0089] Specifically, in this application, the question semantic record can be extracted from the question stem text, option text, and analysis text of the candidate questions. The question semantic record can include question stem object fragments, condition constraint fragments, solution target fragments, option interference fragments, and analysis path fragments. The question stem object fragment represents the object described in the question; the condition constraint fragment represents the known conditions given in the question; the solution target fragment represents the type of result the question requires to be output; the option interference fragment represents distracting expressions in multiple-choice or true / false questions; and the analysis path fragment represents the processing procedures related to the test point operation fragments in the analysis text. After the question semantic record is formed, it is semantically anchored with the test point semantic record. Because the question semantic record includes question stem object fragments, condition constraint fragments, solution target fragments, option interference fragments, and analysis path fragments, it can reduce the bias that occurs when matching only based on question stem keywords.

[0090] Specifically, in this application, semantic anchor alignment refers to determining the correspondence between semantic fragments in the question semantic record and semantic fragments in the test point semantic record. Semantic anchors can be pre-configured as test point definition anchors, condition expression anchors, solution target anchors, problem-solving operation anchors, and error cause association anchors. Test point definition anchors are used to align the question stem object fragment and the test point definition fragment; condition expression anchors are used to align the condition constraint fragment and the test point condition fragment; solution target anchors are used to align the solution target fragment and the test point operation fragment; problem-solving operation anchors are used to align the parsing path fragment and the test point operation fragment; and error cause association anchors are used to align the option interference fragment and the common error cause fragment. When performing semantic anchor alignment between the question semantic record and the test point semantic record, the corresponding fragments are first read according to the above semantic anchors, and then it is determined whether the corresponding fragments have the same test point meaning or hierarchical test point meaning, thereby forming a question matching identifier. The question matching identifier is used to indicate the matching status between candidate questions and test points in the stage preparation path sequence.

[0091] Specifically, in this application, semantic vector encoding can be used as an exemplary implementation of semantic anchor alignment. The semantic vector encoding can be generated by a pre-trained text encoding structure, which can employ an encoder structure including a word fragment embedding layer, a position embedding layer, a multi-layer self-attention encoding layer, and a normalized output layer. The word fragment embedding layer converts text fragments in the question semantic record and the test point semantic record into word fragment representations; the position embedding layer preserves the order of word fragments within the text fragment; the multi-layer self-attention encoding layer establishes associations between word fragments within the same text fragment; and the normalized output layer converts text fragments of different lengths into comparable semantic vector codes. Taking one implementable structure as an example, the encoder structure can employ a multi-layer self-attention encoding layer, each layer including an attention calculation sublayer and a feedforward transformation sublayer. The attention calculation sublayer associates key conditions within the same text fragment with the solution objective, and the feedforward transformation sublayer converts the output of the attention calculation sublayer into a semantic vector code. This semantic vector encoding is only one implementation of semantic anchor alignment; the question type matching record still uses the question matching identifier and candidate questions as the final content.

[0092] Specifically, in this application, if semantic vector encoding is used for semantic anchor alignment, each semantic segment in the question semantic record and the test point semantic record can be encoded separately to form question semantic vector encoding and test point semantic vector encoding. Each element in the question semantic vector encoding can represent the expression intensity of a semantic segment in the question semantic record in the corresponding semantic dimension, and each element in the test point semantic vector encoding can represent the expression intensity of a semantic segment in the test point semantic record in the corresponding semantic dimension. When comparing the question semantic vector encoding and the test point semantic vector encoding, the semantic vector encoding of the corresponding segment can be compared item by item according to the semantic anchor, and the comparison result is then converted into a question matching identifier. In this process, both the question semantic vector encoding and the test point semantic vector encoding originate from the question semantic record and the test point semantic record, and the question matching identifier is written back to the candidate questions and participates in the generation of question type matching records. Therefore, there will be no semantic processing results that are out of sync with the stage preparation path sequence.

[0093] Specifically, in this application, the question matching identifier may include a matching test point identifier, a matching anchor point identifier, a matching status identifier, and a question type adaptation identifier. The matching test point identifier points to test points in the stage preparation path sequence; the matching anchor point identifier records the semantic anchor points that are aligned with the question semantic record and the test point semantic record; the matching status identifier indicates whether a candidate question is suitable as a practice question for that test point; and the question type adaptation identifier records the correspondence between the question type of the candidate question and the test point. After the question matching identifier is formed, it is written into the question type matching record of the corresponding test point along with the candidate question. The question type matching record includes at least a test point identifier, a candidate question, a question matching identifier, and a question type adaptation identifier. After the question type matching record is formed, it participates in the generation of the preparation path recommendation record, enabling each test point in the preparation path recommendation record to be associated with a candidate question that better matches the test point semantic record.

[0094] Specifically, this application uses the function extremum test point as an example to illustrate the generation of question type matching records. After the test point identifier in the stage preparation path sequence points to the function extremum, the read test point semantic record includes the test point definition fragment, the test point condition fragment, and the solution target fragment of the function extremum. The preset question bank index obtains candidate questions based on the test point semantic record. If the question semantic record of a candidate question includes question stem object fragments such as finding the maximum value, finding the minimum value, analyzing the monotonic interval, and determining the extremum point, as well as solution target fragments, then the question semantic record and the function extremum test point semantic record form a correspondence in the test point definition anchor point, condition expression anchor point, and solution target anchor point, thereby generating a question matching identifier. The candidate question carrying the question matching identifier is written into the question type matching record corresponding to the function extremum, and the subsequent preparation path recommendation record can establish a path node association relationship between the function extremum and the candidate question.

[0095] Specifically, in this application, semantic anchor alignment can also utilize error-causing markers to correct the matching status of candidate questions. If the error-causing markers in the test preparation profile record indicate that the test preparation candidate has a concept recognition error under a certain test point, then when performing semantic anchor alignment between the question semantic record and the test point semantic record, priority can be given to the test point definition anchor and option interference fragments; if the error-causing markers indicate that the test preparation candidate has a condition extraction error, then priority can be given to the condition expression anchor and condition constraint fragments. The question matching identifier after being corrected by the error-causing markers is still written into the question type matching record and continues to participate in the generation of the test preparation path recommendation record. Through this processing, the question type matching record can cooperate with the error-causing markers in the test preparation profile record, instead of being generated solely based on the static test point annotations in the question bank index.

[0096] Optionally, a test preparation path recommendation record is generated based on the test preparation path sequence and question type matching record, and the test preparation profile record is updated based on the execution feedback of the test preparation path recommendation record, including: Establish path node associations for candidate questions in the test point and question type matching records in the phased preparation path sequence; Based on the relationship between path nodes, a test preparation path recommendation record is generated. Among them, the stage test preparation path sequence determines the order of test points in the test preparation path recommendation record, and the question type matching record determines the candidate questions corresponding to each test point. Obtain the path execution feedback generated after the recommended test preparation path is executed, update the behavior attention mark, error cause mark, mastery gap mark and review interval mark based on the path execution feedback, and write the updated marks into the test preparation profile record for use in the next test preparation path recommendation cycle.

[0097] Specifically, in this application, the path node association is used to bind the test points in the stage preparation path sequence with the candidate questions in the question type matching record as executable path nodes. The stage preparation path sequence provides the order of test points, the question type matching record provides the candidate questions corresponding to each test point, and the path node association combines these two processing results into path nodes in the preparation path recommendation record. When establishing the path node association, the test point identifier and path position identifier in the stage preparation path sequence are read first, and then the candidate questions corresponding to the same test point identifier are searched in the question type matching record. If there are multiple candidate questions in the question type matching record, the order of the candidate questions can be determined according to the matching status identifier and question type adaptation identifier in the question matching identifier. After the path node association is formed, it will serve as the basis for assembling the nodes in the preparation path recommendation record.

[0098] Specifically, in this application, the test preparation path recommendation record includes the test point order, candidate questions, path node relationships, and recommendation description markers. The test point order is derived from the phased test preparation path sequence, the candidate questions are derived from the question type matching record, the path node relationships are used to connect the test point order and the candidate questions, and the recommendation description markers can be generated jointly by the test point priority status identifier, the pre-correction item, and the question matching identifier. The recommendation description markers can include priority source descriptions, pre-correction reinforcement descriptions, and question type matching descriptions. The priority source description indicates the main source of the test point's entry into the test preparation path recommendation record; the pre-correction reinforcement description indicates whether the test point comes from a pre-correction reinforcement item or is subject to a downgrade identifier; the question type matching description indicates the semantic anchor correspondence between the candidate questions and the test point semantic record. After the test preparation path recommendation record is formed, it can be output to the display interface of the test preparation system for display, or it can be stored in the recommendation history of the test preparation object so that the test preparation profile record can be updated after the execution feedback of the test preparation path recommendation record arrives. The test preparation system's display interface can read the test point order, candidate questions, and recommendation description markers from the test preparation path recommendation record, and display the corresponding path nodes according to the path node association relationship.

[0099] Specifically, in this application, the "stage preparation path sequence determines the order of test points in the test preparation path recommendation record" means that the path position identifiers in the stage preparation path sequence are directly written into the test preparation path recommendation record. The "question type matching record determines the candidate questions corresponding to each test point" means that the candidate questions and question matching identifiers in the question type matching record are written into the corresponding path node in the test preparation path recommendation record. If a test point in the stage preparation path sequence comes from a preceding reinforcement item, then that test point will retain its preceding source identifier in the test preparation path recommendation record; if a test point in the stage preparation path sequence is subject to a downgrade identifier, then that test point will retain its downgrade identifier in the test preparation path recommendation record. Through this writing method, the test preparation path recommendation record not only displays test points and questions, but also retains test point priority profiles, preceding correction items, and the processing sources of the stage preparation path sequence and question type matching record.

[0100] Specifically, in this application, path execution feedback refers to the feedback data generated after the test-taker executes the recommended test-taker path record. Path execution feedback can include the test point learning completion status, candidate question answering status, explanation viewing status, correction completion status, and review time status. The test point learning completion status indicates whether the test-taker has completed the learning task for the corresponding test point; the candidate question answering status indicates whether the test-taker has answered the corresponding candidate question; the explanation viewing status indicates whether the test-taker has viewed the explanation of the candidate question; the correction completion status indicates whether the test-taker has corrected incorrect candidate questions; and the review time status indicates the study time spent by the test-taker on the corresponding test point or candidate question. After the path execution feedback is generated, it is matched with the path node associations in the recommended test-taker path record to determine the execution status corresponding to each path node.

[0101] Specifically, in this application, when updating the behavior attention marker based on path execution feedback, the test point corresponding to the path execution feedback can be determined first based on the path node association relationship, and then the review duration status, analysis viewing status, and test point learning completion status can be converted into a new behavior attention marker. The new behavior attention marker continues to be written into the exam preparation profile record and participates in the formation of the test point status mapping record in the next exam preparation path recommendation cycle. When updating the error cause pointing marker based on path execution feedback, new error cause sources can be identified based on the candidate question answering status and correction completion status, and the new error cause sources can be written into the error cause pointing marker. The new error cause pointing marker will participate in the setting of test point priority status identifiers and the correction of question matching identifiers in the next exam preparation path recommendation cycle. When updating the mastery gap marker based on path execution feedback, it can be determined whether the mastery status of the test point has changed based on the test point learning completion status, candidate question answering status, and correction completion status, and the changed mastery status can be written into the mastery gap marker. The new mastery gap marker will participate in the pre-readiness verification in the next exam preparation path recommendation cycle. When updating the review interval marker based on path execution feedback, the execution time of the recommended study path can be written to the historical review time record, and then converted from the historical review time record into a new review interval marker. The new review interval marker will participate in the setting of the test point priority status indicator in the next study path recommendation cycle.

[0102] Specifically, in this application, the updated behavior focus markers, error cause pointing markers, mastery gap markers, and review interval markers need to be written into the same test preparation profile record. During writing, the data identifiers formed by the test preparation object identifier, target exam configuration identifier, and test preparation path recommendation cycle identifier can be reused, and feedback source identifiers can be configured for the updated behavior focus markers, error cause pointing markers, mastery gap markers, and review interval markers. The feedback source identifier is used to indicate which test preparation path recommendation record's path execution feedback the updated markers originate from. After the updated test preparation profile record is formed, it will re-enter the acquisition step in the next test preparation path recommendation cycle and serve as the basis for generating the test point priority profile. Therefore, the execution feedback of the test preparation path recommendation record is not merely stored as a completed state, but is transformed into a test preparation profile record that can re-influence the priority of candidate test points, pre-readiness checks, and question type matching.

[0103] Specifically, this application uses two exam points—function extrema and proof of derivatives using inequalities—as examples to illustrate the relationship between the exam preparation path recommendation record and the path execution feedback. If function monotonicity is located before proof of derivatives using inequalities in the stage exam preparation path sequence, and the question type matching record matches several candidate questions for function monotonicity and several candidate questions for proof of derivatives using inequalities, then the exam preparation path recommendation record will form corresponding path nodes according to the order of exam points in the stage exam preparation path sequence. After the exam taker executes the exam preparation path recommendation record, if the answer status of the candidate questions for function monotonicity shows "passed," the correction completion status shows "completed," and the review time status shows "learning completed," then the path execution feedback will update the mastery gap marker and review interval marker corresponding to function monotonicity. If the answer status of the candidate questions for proof of derivatives using inequalities still shows "failed," then the path execution feedback will update the error reason marker and mastery gap marker corresponding to proof of derivatives using inequalities. When the next preparation path recommendation cycle arrives, the updated preparation profile record will be re-involved in setting the test point priority status indicator, so that subsequent preparation path recommendation records can reflect the status changes brought about by the feedback from the previous round of path execution.

[0104] Specifically, in this application, the test preparation path recommendation record may also include a path version identifier. The path version identifier can be formed by combining the test preparation path recommendation cycle identifier and the generation order of the test preparation path recommendation records. After the path version identifier is written into the test preparation path recommendation record, the path execution feedback will return with the path version identifier of the same test preparation path recommendation record. Based on the path version identifier, the path execution feedback can be matched with the corresponding test preparation path recommendation record, preventing path execution feedback generated in different test preparation path recommendation cycles from being mixed into the same test preparation profile record. After the path version identifier participates in the feedback write-back, it will enter the next test preparation path recommendation cycle along with the updated test preparation profile record, enabling the next test preparation path recommendation cycle to trace the source of the updated behavior attention markers, error cause pointing markers, mastery gap markers, and review interval markers.

[0105] Specifically, in this application, through the connection between the phased preparation path sequence, question type matching record, preparation path recommendation record, and path execution feedback, the preparation path recommendation method can transform the preparation profile record before recommendation into the preparation profile record after recommendation. The phased preparation path sequence is responsible for determining the order of test points, the question type matching record is responsible for determining the candidate questions corresponding to each test point, the preparation path recommendation record is responsible for converting the test point order and candidate questions into an executable path, and the path execution feedback is responsible for rewriting the execution status into behavior attention markers, error cause markers, mastery gap markers, and review interval markers. The updated preparation profile record will participate again in the generation of test point priority profiles in the next preparation path recommendation cycle, thereby enabling the preparation path recommendation record to be continuously adjusted according to the actual execution status of the test taker.

[0106] In a specific application scenario, this application can be applied to the second-round review of senior high school mathematics. The test-taker can be a senior high school science student preparing for the exam. The target exam scope record can correspond to the exam scope of high school mathematics, such as functions and derivatives, solid geometry, sequences, trigonometric functions, probability and statistics, and analytic geometry. The current preparation stage record can correspond to the second-round review stage. The test-taking path recommendation cycle can be configured to generate a test-taking path recommendation record once a day. In this specific application scenario, the test-taking profile record, the current preparation stage record, the target exam scope record, and the test point dependency graph are all associated with the same test-taking object identifier. This allows the learning behavior records, answer feedback records, mastery status records, and historical review time records formed by the test-taking object within a test-taking path recommendation cycle to be continuously converted into behavior attention markers, error reason markers, mastery gap markers, and review interval markers, and to continue participating in the test point priority status marker setting for candidate test points. The test-taking object identifier can be written into the test-taking path recommendation record along with the test-taking path recommendation cycle, so that the path execution feedback generated by the subsequent test-taking path recommendation record can be written back to the same test-taking profile record.

[0107] Specifically, this application uses function extrema as a candidate test point for explanation. The learning behavior record of test takers regarding function extrema test points within the most recent statistical period can include actions such as entering the function extrema explanation page, viewing function extrema error analysis, actively searching for function extrema related questions, and collecting function extrema error questions. The most recent statistical period can be formed by reading backwards from the test preparation path recommendation period. For example, when generating test preparation path recommendation records daily, learning behavior records generated within several days prior to the current recommendation time can be read. When performing structured transformation on the above learning behavior records, the number of times entering the function extrema explanation page can be converted into browsing frequency status, the cumulative time spent viewing function extrema error analysis can be converted into dwell time status, the number of times actively searching for function extrema related questions can be converted into search activity status, and the record of collecting function extrema error questions can be converted into collection trigger status. Browsing frequency status, dwell time status, search activity status, and collection trigger status together form the behavioral attention marker corresponding to function extrema. The behavioral attention marker is then written into the test preparation profile record and mapped to the function extrema candidate test point when the test point priority profile is generated.

[0108] Specifically, in this application, behavioral attention markers can be converted into learning behavior weights through normalization. The learning behavior weight is denoted by a capital letter B followed by a test point number subscript, representing the candidate's active attention to a particular candidate test point; the test point number subscript indicates the candidate test point's position in the target exam scope record. The learning behavior weight can be generated based on browsing frequency, dwell time, search activity, and collection trigger status. Taking a function extreme value test point as an example, if the normalized value corresponding to the browsing frequency status is 0.6, the normalized value corresponding to the dwell time status is 0.75, the normalized value corresponding to the search activity status is 0.6, and the normalized value corresponding to the collection trigger status is 0.85, then these four normalized values ​​can be combined into the learning behavior weight corresponding to the function extreme value by averaging or using a preset ratio. This learning behavior weight is not a separately stored statistical indicator but is written into the test point status mapping record and continues to participate in the generation of the test point priority status identifier for function extreme values. For the same candidate test point, the behavioral attention markers that are re-formed in the subsequent test preparation path recommendation cycle are still converted into learning behavior weights in the same way as described above, so that the learning behavior weights can be updated as the learning behavior records change.

[0109] Specifically, in this application, the answer feedback record can be further converted into error cause pointing tags. Taking the function extremum test point as an example, if the candidate's recent answer feedback record for function extremum questions shows a 48% accuracy rate, and the errors mainly correspond to conceptual misunderstandings, calculation errors, and deviations in problem-solving approaches, then the answer feedback record is first assigned to the function extremum test point according to the question identifier, and then converted into error cause pointing tags according to the source of the error. Error cause pointing tags are used to describe the sources of errors affecting the answer results in the function extremum test point. The mastery status record can be generated based on the historical accuracy rate of function extremum questions, continuous answering performance, and performance after correction, and then converted into mastery gap tags. Taking a 48% accuracy rate for the function extremum test point as an example, the mastery status value can be converted to 0.48, and the mastery weight of the knowledge point corresponding to the mastery gap tag can be set to one minus the mastery status value, i.e., 0.52. The knowledge point mastery weight here is represented by a capital letter M followed by the test point number subscript. This indicates the need for reinforcement of the candidate test point in terms of mastery. The higher the knowledge point mastery weight, the more priority the candidate test point should be given in the current preparation profile. Errors in concept understanding, calculation mistakes, and deviations in problem-solving approaches are all included in the generation of subsequent question type matching records through error cause pointing tags, avoiding the error cause pointing tags only remaining at the statistical level of answer feedback records.

[0110] Specifically, in this application, historical review time records can be converted into review interval markers, and further, time decay weights can be generated. The time decay weight is denoted by a capital letter D followed by the test point number subscript, representing the impact of the interval between the candidate test point and the last effective learning or practice. The time decay weight can be expressed in exponential decay form as: D_i = e^(-λ·Δt_i). Where D_i represents the time decay weight of the i-th candidate test point; e represents the natural constant; λ represents the time decay coefficient, which can be pre-configured based on the current preparation stage records and the subject review rhythm; Δt_i represents the number of days between the i-th candidate test point and the last effective learning or practice. Taking the test point related to function monotonicity as an example, where the last practice was 18 days ago, when λ is 0.02, D_i equals e to the power of -0.36, corresponding to approximately 0.70. When the preparation system smooths the time decay weight according to the second round of review stage corresponding to the current preparation stage records, this time decay weight can be configured to a state value close to 0.65. The time decay weight is written into the review interval marker and participates in the generation of the test point priority status identifier through the review interval marker. The smoothing process here can be implemented using a preset interval mapping method. For example, the time decay weight can be mapped to the review interval status interval allowed in the current preparation stage, so that the review interval marker can participate in the calculation on the same scale as the learning behavior weight, knowledge point mastery weight, and personalized adjustment weight.

[0111] Specifically, in this application, the personalized adjustment weight is denoted by the uppercase letter P followed by the test point number subscript, used to represent the adjustment effect of the target exam configuration on candidate test points. The target exam configuration may include data such as target institution level, weak module configuration, target score range configuration, and teacher intervention markers. The target institution level, weak module configuration, target score range configuration, and teacher intervention markers are all subordinate data sources of the target exam configuration, used to enable the target exam configuration to form personalized adjustment weights for candidate test points. Taking the test candidate marking the function and derivative module as a weak module as an example, function extrema, function monotonicity, and proof of derivative application inequalities can all obtain higher personalized adjustment weights. After the personalized adjustment weights are written into the test point status mapping record, they participate in the generation of test point priority status identifiers together with learning behavior weights, knowledge point mastery weights, and time decay weights. If the subsequent path execution feedback shows that the mastery gap marker of the function and derivative module has been reduced, the target exam configuration can still keep the weak module configuration unchanged, while the test candidate profile record will reduce the knowledge point mastery weight of some candidate test points in the function and derivative module through the new mastery gap marker.

[0112] Specifically, in this application, multi-dimensional priority scoring can be used as a concrete implementation of the priority status identifier for test points. For the i-th candidate test point, its comprehensive priority score can be expressed as: S_i = w_1·B_i + w_2·M_i + w_3·D_i + w_4·P_i. Where S_i represents the comprehensive priority score of the i-th candidate test point; B_i represents the learning behavior weight of the i-th candidate test point; M_i represents the knowledge point mastery weight of the i-th candidate test point; D_i represents the time decay weight of the i-th candidate test point; P_i represents the personalized adjustment weight of the i-th candidate test point; w_1 represents the stage weight coefficient corresponding to the learning behavior weight; w_2 represents the stage weight coefficient corresponding to the knowledge point mastery weight; w_3 represents the stage weight coefficient corresponding to the time decay weight; and w_4 represents the stage weight coefficient corresponding to the personalized adjustment weight. The above four stage weight coefficients are determined by the current preparation stage record, and the sum of the four stage weight coefficients is one, enabling the comprehensive priority score to express the priority status of the candidate test point on the same scale. After the comprehensive priority score is written to the test center priority status flag, the test center priority status flag is then written to the test center priority profile, so that the test center priority profile can be read by the pre-ready verification.

[0113] Specifically, in this application, when the current preparation stage is recorded as the second round of review, the stage weight coefficients can be configured as follows: a stage weight coefficient of 0.25 corresponding to the learning behavior weight, a stage weight coefficient of 0.30 corresponding to the knowledge point mastery weight, a stage weight coefficient of 0.25 corresponding to the time decay weight, and a stage weight coefficient of 0.20 corresponding to the personalized adjustment weight. Taking the function extreme value test point as an example, if the learning behavior weight is 0.70, the knowledge point mastery weight is 0.52, the time decay weight is 0.65, and the personalized adjustment weight is 0.85, then the comprehensive priority score of the function extreme value test point is 0.25 multiplied by 0.70, plus 0.30 multiplied by 0.52, plus 0.25 multiplied by 0.65, plus 0.20 multiplied by 0.85, resulting in approximately 0.66. This comprehensive priority score will be converted into a test point priority status identifier corresponding to the function extreme value and written into the test point priority profile. Taking the exam topic of proving inequalities using derivatives as an example, if the weight of learning behavior is 0.40, the weight of knowledge point mastery is 0.35, the weight of time decay is 0.80, and the weight of personalized adjustment is 0.70, then its comprehensive priority score is 0.545. This comprehensive priority score is also converted into the exam topic priority status identifier corresponding to the proof of inequalities using derivatives and written into the exam topic priority profile. The exam topic priority status identifiers corresponding to function extrema and the exam topic priority status identifiers corresponding to the proof of inequalities using derivatives will continue to participate in the pre-readiness check.

[0114] Specifically, in this application, the test point dependency graph can be stored in the form of a directed graph. Each test point in the test point dependency graph can be represented as a node in the graph, and the prior learning relationships between test points can be represented as directed edges in the graph. Directed edges point from a prior test point to a candidate test point that depends on that prior test point. For the i-th candidate test point, its prior test point chain can be represented as Pre(K_i) = {K_j | K_j → K_i}. Here, K_i represents the i-th candidate test point; K_j represents a prior test point of K_i; the symbol K_j → K_i indicates that K_j is a prior test point of K_i; Pre(K_i) represents the prior test point chain composed of all prior test points that have a direct prior relationship with K_i. Taking the test point of proving inequalities using derivatives as an example, its prior test point chain can include basic derivative operations, function monotonicity, and basic properties of inequalities. After the prerequisite test point chain is written into the test point dependency graph, it is matched with the mastery gap marker in the test preparation profile record during the prerequisite readiness check. The directed edges in the test point dependency graph can also carry the prerequisite relationship type, which is used to indicate whether the prerequisite test point and the candidate test point have a necessary basic relationship, a direct successor relationship, or a same-level support relationship.

[0115] Specifically, in this application, the pre-examination readiness verification can be implemented using a mastery state threshold as an example. For any pre-examination point K_j in the pre-examination point chain, its mastery state is denoted as A_j, where A_j represents the candidate's mastery level of the pre-examination point K_j. A_j can be obtained based on the mastery state record, answer feedback record, and the answer state transition after correction. The pre-examination readiness condition can be expressed as A_j ≥ T_pre, where T_pre represents the pre-examination readiness threshold, which can be pre-configured based on the current preparation stage record. If each pre-examination point K_j in the pre-examination point chain satisfies A_j ≥ T_pre, the corresponding candidate test point retains its original test point priority state identifier; if any pre-examination point K_j in the pre-examination point chain does not satisfy A_j ≥ T_pre, the corresponding candidate test point is written with a downgrade identifier, and the pre-examination point K_j is written into the pre-examination reinforcement entry. Taking the proof of inequalities using derivatives as an example, if the mastery level of basic derivative operations is 0.75, the mastery level of function monotonicity is 0.55, and the mastery level of basic inequality properties is 0.82, while the prerequisite readiness threshold is 0.60, then function monotonicity does not meet the prerequisite readiness condition. Function monotonicity will be added to the prerequisite reinforcement entry, and the proof of inequalities using derivatives will be added to the downgraded entry. The downgraded entry and the prerequisite reinforcement entry are combined to form the prerequisite correction entry for the proof of inequalities using derivatives.

[0116] Specifically, in this application, the downgrade flag can correct the overall priority score of candidate test points. The correction formula can be expressed as: S_i′ = α·S_i. Where S_i′ represents the overall priority score of the i-th candidate test point after pre-correction; α represents the downgrade coefficient, which is between zero and one, used to indicate the degree of suppression on the candidate test point's ranking position when the pre-readiness check fails; S_i represents the original overall priority score of the i-th candidate test point. Taking the test point on proving inequalities using derivatives as an example, its original overall priority score is 0.545. When the downgrade coefficient is configured to 0.3, its overall priority score after pre-correction is 0.164. This pre-corrected overall priority score will be written into the pre-correction entry along with the downgrade flag and will be applied to the path position of the proof of inequalities using derivatives during stage adaptation. The downgrade coefficient can be configured according to the current preparation stage. For example, a lower downgrade coefficient can be configured in the basic coverage stage to strengthen the basics, a medium downgrade coefficient can be configured in the topic strengthening stage to balance topic advancement and pre-exam reinforcement, and the downgrade coefficient can be adjusted in combination with the path length control indicator in the sprint stage.

[0117] Specifically, in this application, the preceding reinforcement entries can continue to undergo upstream preceding test point chain verification. Taking the function monotonicity as an example of a preceding reinforcement entry, if the test point dependency graph shows that the function monotonicity also depends on the basic meaning of the derivative, then the basic meaning of the derivative constitutes the upstream preceding test point of the function monotonicity. If the mastery of the basic meaning of the derivative has not reached the preceding readiness threshold, the basic meaning of the derivative will be added to the preceding reinforcement entry and arranged before the function monotonicity. If the mastery of the basic meaning of the derivative reaches the preceding readiness threshold, the function monotonicity can be directly configured as an insertable state. The insertable state is used to indicate that the preceding reinforcement entry has completed the upstream preceding test point chain preceding readiness verification and can be arranged before the corresponding candidate test point during stage adaptation. Through upstream preceding test point chain verification, the preceding reinforcement entries not only include direct preceding test points, but also upstream preceding test points that have a continuing dependency relationship with the direct preceding test points.

[0118] Specifically, in this application, stage adaptation can be achieved through stage weight coefficients, stage task boundaries, and path length control identifiers. When the current preparation stage is recorded as the first round of review, the stage weight coefficients can be configured as follows: 0.15 for the learning behavior weight, 0.40 for the knowledge point mastery weight, 0.25 for the time decay weight, and 0.20 for the personalized adjustment weight. The stage task boundaries are biased towards basic and medium-difficulty test points, and the path length control identifier can be configured as a relatively long path. When the current preparation stage is recorded as the second round of review, the stage weight coefficients can be configured as 0.25, 0.30, 0.25, and 0.20. The stage task boundaries are biased towards topical test points and medium-to-high difficulty test points, and the path length control identifier can be configured as a medium-length path. When the current preparation stage is recorded as the sprint stage, the stage weight coefficient can be configured to 0.30, 0.20, 0.35, and 0.15. The stage task boundaries are biased towards high-frequency test points and error-prone question types, and the path length control identifier can be configured to a shorter path. The above stage weight coefficients and path length control identifiers are all exemplary configurations and can be adjusted according to the target exam configuration and preparation calendar status in specific applications. The stage task boundaries and path length control identifiers are written to the current preparation stage record and will continue to be read when generating the stage preparation path sequence.

[0119] Specifically, in this application, the sequence of stage preparation paths can be arranged according to the comprehensive priority score after pre-correction, while being constrained by the insertability of pre-reinforcement items. Taking the second round of review for senior high school mathematics as an example, when the current preparation stage is recorded as the second round of review, the stage task boundary can be selected from topics such as functions and derivatives, solid geometry, and sequences. If the comprehensive priority score for function extrema is approximately 0.66, the comprehensive priority score for determining the perpendicularity of lines and planes in solid geometry is approximately 0.61, function monotonicity is written into the pre-reinforcement item, and the comprehensive priority score for proving derivatives using inequalities after downgrading is approximately 0.164, then the stage adaptation can generate the following stage preparation path sequence: function extrema, determining the perpendicularity of lines and planes in solid geometry, function monotonicity, finding the general term formula of sequences, trigonometric function graph transformation, proving derivatives using inequalities, probability distribution sequences, and comprehensive analysis of parabolas in analytic geometry. In this stage of the preparation path sequence, function monotonicity precedes the proof of derivative application inequalities, demonstrating the role of pre-reinforcement items in the stage preparation path sequence; the proof of derivative application inequalities follows function monotonicity, demonstrating the role of degradation markers in the path positions of candidate test points. After the stage preparation path sequence is formed, each test point retains its corresponding test point identifier and path position identifier for subsequent reading of test point semantic records and generation of question type matching records.

[0120] Specifically, in this application, the semantic records of questions and test points can be aligned using a large language model for educational texts. This large language model can employ a text encoding structure based on a Transformer encoder. Instead of using convolutional operations to form local sliding window features, this structure establishes global connections between different word segments within a text fragment based on a self-attention matrix. The large language model for educational texts may include a word segment embedding layer, a position embedding layer, a multi-head self-attention layer, a feedforward transform layer, a normalization layer, and a semantic projection layer. The word fragment embedding layer converts word fragments from the question stem, option text, parsing text, test point definitions, and related concepts into word fragment vectors. The positional embedding layer adds the order information of the word fragments in the text to the word fragment vectors. The multi-head self-attention layer calculates the association strength between different word fragments, aligning the conditional expressions, solution objectives, and test point definitions in the question within the same semantic space. The feedforward transformation layer performs a non-linear transformation on the semantic representation output by the multi-head self-attention layer. The normalization layer stabilizes the semantic representation under different text lengths. The semantic projection layer converts the question semantic records and test point semantic records into question semantic vectors and test point semantic vectors of the same dimension. The question semantic vectors and test point semantic vectors then participate in semantic similarity calculation and are used to generate question matching identifiers.

[0121] Specifically, in this application, the multi-head self-attention layer can be represented as follows: Attention(Q,K,V) = Softmax((QK^T) / sqrt(d))V. Here, Q represents the query matrix, which is obtained by a trainable linear transformation of the word vectors in the current text segment; K represents the key matrix, which is obtained by another trainable linear transformation of the word vectors in the same text segment; V represents the value matrix, which is obtained by yet another trainable linear transformation of the word vectors in the same text segment; QK^T represents the similarity matrix after transposing the query matrix and the key matrix; d represents the vector dimension in each attention head; sqrt(d) represents the square root term for scaling the similarity matrix; Softmax represents converting the similarity matrix into an attention weight matrix; and Attention(Q,K,V) represents the semantic representation obtained by weighting the value matrix using the attention weight matrix. For the question semantic record, the semantic representation can express the relationship between the question stem object fragment, condition constraint fragment, solution target fragment, option interference fragment, and analysis path fragment; for the test point semantic record, the semantic representation can express the relationship between the test point definition fragment, test point condition fragment, test point operation fragment, related concept fragment, and common error cause fragment. Thus, the semantic representation output by the multi-head self-attention layer can be further converted into question semantic vectors and test point semantic vectors by the semantic projection layer.

[0122] Specifically, in this application, the educational text large language model can adopt a multi-layer Transformer encoder structure as an example, such as a twelve-layer Transformer encoder. Each Transformer encoder layer includes a multi-head self-attention layer, a feedforward transformation layer, and a normalization layer. The output of each Transformer encoder layer serves as the input to the next Transformer encoder layer. This number of layers is only an example of implementation and is not a limitation. After the question semantic record is input into the educational text large language model, the semantic projection layer outputs the question semantic vector; after the test point semantic record is input into the educational text large language model, the semantic projection layer outputs the test point semantic vector. The question semantic vector is denoted as V_q, and the test point semantic vector is denoted as V_k. Wherein, V_q represents the semantic representation of the candidate question, and V_k represents the semantic representation of a certain test point in the stage preparation path sequence. Each dimension of the question semantic vector and the test point semantic vector is used to represent the strength of an abstract semantic dimension, such as the semantic dimensions of condition expression, target type, problem-solving operation, related concepts, or error sources. After the question semantic vector and the test point semantic vector are generated, they will enter the semantic similarity calculation to form a question matching identifier.

[0123] Specifically, in this application, the semantic similarity between the question semantic record and the test point semantic record can be represented by the cosine similarity formula: Sim(V_q,V_k) = (V_q · V_k) / (||V_q|| · ||V_k||). Where Sim(V_q,V_k) represents the semantic similarity between the candidate question and the test point; V_q represents the question semantic vector; V_k represents the test point semantic vector; V_q · V_k represents the dot product of the question semantic vector and the test point semantic vector; ||V_q|| represents the length of the question semantic vector; and ||V_k|| represents the length of the test point semantic vector. The closer the semantic similarity is to one, the closer the question semantic record of the candidate question is to the test point semantic record. A semantic matching threshold T_sim can be pre-configured, where T_sim represents the minimum semantic similarity required for a candidate question to be included in the question type matching record. If Sim(V_q,V_k) is greater than or equal to T_sim, a question matching identifier is generated for the candidate question, and the candidate question carrying the question matching identifier is written into the question type matching record; if Sim(V_q,V_k) is less than T_sim, the candidate question is not written into the question type matching record corresponding to that test point. After the question type matching record is formed, it will continue to participate in the generation of the preparation path recommendation record.

[0124] Specifically, in this application, the large language model for educational texts can be trained using labeled samples of questions and test points. Training samples can include positive and negative samples. Positive samples are those where the semantic record of the question matches the corresponding semantic record of the test point, while negative samples are those where the semantic record of the question does not match the semantic record of a non-corresponding test point. During training, the semantic vectors of the question and the semantic vectors of the test points can be concatenated and input into the semantic matching classification layer, and a binary cross-entropy loss function can be used for training. This loss function can be expressed as: L = -[y·log(p) + (1-y)·log(1-p)]. Here, L represents the training loss of the semantic matching classification layer; y represents the sample matching label, where y equals one indicating a match between the semantic record of the question and the semantic record of the test point, and y equals zero indicating a mismatch; p represents the matching probability output by the semantic matching classification layer; and log represents the natural logarithm. The matching probability p can be obtained by concatenating the semantic vector of the question, the semantic vector of the test point, and their difference vectors, followed by a linear transformation and activation transformation. This training process aims to establish a relatively stable alignment between the question semantic records and the test point semantic records in the semantic space. After training, the educational text large language model can generate question semantic vectors and test point semantic vectors in the test preparation path recommendation method. The question semantic vectors and test point semantic vectors are then used for semantic similarity calculation, which affects the question matching identifier.

[0125] Specifically, in this application, taking the function extremum test point as an example, the test point semantic record can include semantic fragments such as the definition of function extremum, determining stationary points after differentiation, and determining extremum points by combining monotonic intervals. The candidate questions in the question bank can include Question 1, Question 2, and Question 3; the semantic stem of Question 1 is to find the maximum and minimum values, the semantic stem of Question 2 is to analyze monotonicity and determine extremum points, and the semantic stem of Question 3 is to find the peak and valley values. After inputting Question 1, Question 2, and Question 3 into the educational text large language model, their respective question semantic vectors can be obtained; after inputting the test point semantic record of function extremum into the educational text large language model, the test point semantic vector corresponding to the function extremum can be obtained. If the semantic similarity between Question 1 and the function extremum is 0.92, the semantic similarity between Question 2 and the function extremum is 0.88, and the semantic similarity between Question 3 and the function extremum is 0.85, and the semantic matching threshold is 0.75, then Question 1, Question 2, and Question 3 will all generate question matching identifiers and be written into the question type matching record corresponding to the function extremum. The matching record for this question type is then linked to the function extrema in the stage preparation path sequence to establish a path node relationship.

[0126] Specifically, in this application, error-causing markers can also participate in the generation of question type matching records. Taking the function extremum test point as an example, if the error-causing markers indicate that the test taker mainly has conceptual misunderstanding errors in the function extremum test point, then when aligning the semantic anchor points of the question semantic record and the test point semantic record, the comparison ranking of the test point definition anchor point and the option interference fragment can be improved, so that candidate questions focusing on conceptual analysis have a higher ranking in the question type matching record. If the error-causing markers indicate that the test taker mainly has calculation process errors, then the comparison ranking of the analytical path fragment and the test point operation fragment can be improved, so that candidate questions focusing on differentiation operations and process verification can be placed in a higher position. If the error-causing markers indicate that the test taker mainly has a problem-solving approach deviation, then the comparison ranking of the solution target fragment and the analytical path fragment can be improved, so that candidate questions focusing on problem-solving path selection can be placed in a higher position. Thus, the question type matching record not only reflects the semantic similarity between candidate questions and test point semantic records, but also reflects the moderating effect of error-causing markers on question arrangement in the test taker profile record.

[0127] Specifically, in this application, the preparation path can be generated through path node associations. For each test point K_i in the stage preparation path sequence, the candidate question Q_i corresponding to the test point K_i is read from the question type matching record, and a path node association (K_i, Q_i) is established. The preparation path recommendation record can be represented as Path = [(K_1, Q_1), (K_2, Q_2), ..., (K_N, Q_N)]. Wherein, Path represents the preparation path recommendation record; K_1 to K_N represent the test points arranged in order in the stage preparation path sequence; Q_1 to Q_N represent the candidate questions corresponding to K_1 to K_N respectively; N represents the number of test points in the current preparation path recommendation record, and N is determined by the path length control identifier in the current preparation stage record. Taking the second round of review for senior high school mathematics as an example, when the path length control identifier is configured to eight test points, the recommended preparation path record can include function extrema and their candidate questions, the determination of line-plane perpendicularity in solid geometry and its candidate questions, the monotonicity of functions and its candidate questions, the method for finding the general term formula of a sequence and its candidate questions, the transformation of trigonometric function graphs and its candidate questions, the proof of inequalities applied to derivatives and its candidate questions, probability distribution sequences and their candidate questions, and the comprehensive analysis of parabolas in analytic geometry and its candidate questions. The function monotonicity in this recommended preparation path record is located before the proof of inequalities applied to derivatives, originating from the insertable state of the preceding reinforcement entries; the proof of inequalities applied to derivatives retains a downgrade identifier, originating from the preceding readiness check. After the recommended preparation path record is formed, the path node association is used to bind each test point to its candidate questions for subsequent collection of path execution feedback.

[0128] Specifically, in this application, the path execution feedback of the exam preparation path recommendation record can continue to update the exam preparation profile record. If the exam candidate completes the candidate question corresponding to the function monotonicity and answers it, the candidate question answering status, correction completion status, and review duration status in the path execution feedback will be converted into new mastery gap markers and new review interval markers. The new mastery gap markers can improve the mastery status of function monotonicity in the next exam preparation path recommendation cycle, and the new review interval markers can prevent function monotonicity from being repeatedly prioritized due to excessively long review intervals in the short term. If the exam candidate still fails to answer the candidate question corresponding to the derivative application inequality proof, the path execution feedback will enhance the error cause pointing marker and mastery gap marker corresponding to the derivative application inequality proof, so that the exam point will continue to participate in the exam point priority profile generation and pre-examination readiness verification in the next exam preparation path recommendation cycle. In this way, a continuously updated data processing chain is formed between the exam preparation path recommendation record, path execution feedback, and exam preparation profile record.

[0129] Specifically, in this application, the aforementioned specific application scenarios are only used to illustrate how the exam preparation path recommendation method is implemented in the multi-dimensional priority scoring, knowledge graph pre-dependency verification, stage adaptation strategy, question semantic fusion, and exam preparation path generation process in the disclosure document. The exam preparation target, exam point name, weight value, threshold value, number of questions, and path length can all be configured based on the target exam scope record, the current exam preparation stage record, and the exam preparation profile record. Even if the high school mathematics function and derivative exam preparation scenario is replaced with an English grammar exam preparation scenario, the same data processing method can still be used to write English grammar exam points, grammar question types, grammar error causes, and grammar pre-relationships into the exam preparation profile record, exam point dependency graph, exam point priority profile, pre-correction entries, stage exam preparation path sequence, question type matching record, and exam preparation path recommendation record. In the English grammar exam preparation scenario, the English grammar exam points, grammar question types, grammar error causes, and grammar pre-relationships correspond to the candidate exam points, candidate questions, error cause pointing markers, and pre-exam point chains in the mathematics exam preparation scenario, respectively, without changing the data processing order of the exam preparation path recommendation method.

[0130] like Figure 2 As shown, this is an embodiment of a test preparation path recommendation device according to this application, comprising: The periodic data acquisition module is used to acquire the preparation profile record, current preparation stage record, target exam scope record, and test point dependency graph corresponding to the target exam scope record when the recommended preparation path period is reached. The priority profile generation module is used to set the priority status identifier of candidate test points in the test point dependency graph based on the test preparation profile records, and generate test point priority profiles. The pre-correction generation module is used to perform pre-readiness verification on the priority profile of test points based on the test point dependency graph, and generate pre-correction entries for candidate test points that do not meet the pre-readiness conditions. The stage path generation module is used to adapt the test point priority profile and the pre-correction items to the current test preparation stage records, and generate a stage test preparation path sequence. The question type matching and generation module is used to obtain the question semantic records according to the stage preparation path sequence, semantically align the question semantic records with the test point semantic records, and generate question type matching records. The recommendation feedback update module is used to generate a preparation path recommendation record based on the stage preparation path sequence and question type matching record, and update the preparation profile record based on the execution feedback of the preparation path recommendation record.

[0131] like Figure 3 As shown, an electronic device according to an embodiment of this application includes a processor and a memory. The memory stores computer instructions, and the processor is used to read and execute the computer instructions to implement the above-mentioned alternative path recommendation method.

[0132] like Figure 4 As shown, this is a computer-readable storage medium according to an embodiment of the present application. The computer-readable storage medium stores computer instructions, which, when executed by a processor, implement the above-mentioned alternative path recommendation method.

[0133] like Figure 5 As shown, this is an embodiment of a test preparation path recommendation system, which includes a front-end test preparation learning device and a cloud-based test preparation path recommendation platform. The front-end exam preparation learning device is used to: collect exam preparation process data from the exam candidates, send the exam preparation process data to the cloud-based exam preparation path recommendation platform, and display the exam preparation path recommendation records returned by the cloud-based exam preparation path recommendation platform; the front-end exam preparation learning device is also used to collect path execution feedback generated after the exam candidates execute the exam preparation path recommendation records, and send the path execution feedback to the cloud-based exam preparation path recommendation platform; The cloud-based exam preparation path recommendation platform is used for: maintaining exam preparation profile records based on exam preparation process data; maintaining records for the current exam preparation stage based on the exam preparation path recommendation cycle; and maintaining a test point dependency graph based on the target exam scope record. It generates test point priority profiles based on the exam preparation profile records and test point dependency graphs; performs pre-readiness checks on the test point priority profiles based on the test point dependency graphs, generates pre-correction items, and performs stage adaptation on the test point priority profiles and pre-correction items based on the current exam preparation stage records, generating a stage-specific exam preparation path sequence. It performs semantic alignment between question semantic records and test point semantic records based on the stage-specific exam preparation path sequence, generates question type matching records, and generates exam preparation path recommendation records based on the stage-specific exam preparation path sequence and question type matching records. Finally, it updates the exam preparation profile records based on path execution feedback, enabling the updated exam preparation profile records to participate in the next exam preparation path recommendation cycle.

[0134] The above Figures 2-5 For an exemplary description, please refer to the above. Figure 1 This will not be elaborated upon here.

Claims

1. A method for recommending test preparation paths, characterized in that, include: When the recommended preparation path cycle is reached, obtain the preparation profile record, current preparation stage record, target exam scope record, and test point dependency graph corresponding to the target exam scope record of the candidate. Based on the test preparation profile records, test point priority status identifiers are set for candidate test points in the test point dependency graph to generate test point priority profiles. Based on the test point dependency graph, a pre-readiness check is performed on the test point priority profile, and pre-correction entries are generated for candidate test points that do not meet the pre-readiness conditions. Based on the current preparation stage records, the priority profiles of test points and the pre-correction items are adapted to the stage to generate a stage preparation path sequence. Based on the sequence of preparation paths, obtain the semantic records of questions, align the semantic records of questions with the semantic records of test points, and generate question type matching records. Based on the sequence of preparation paths and the matching records of question types, a preparation path recommendation record is generated, and the preparation profile record is updated based on the execution feedback of the preparation path recommendation record.

2. The test preparation path recommendation method according to claim 1, characterized in that, Obtain the candidate's preparation profile record, current preparation stage record, target exam scope record, and exam point dependency graph corresponding to the target exam scope record, including: Read the learning behavior records of the test takers from the test preparation system and convert the learning behavior records into behavior attention tags; Read the test-taking feedback records of the test takers during the question-answering process from the test preparation system, and convert the feedback records into error reason markers; Read the mastery status records of test takers in the test point dimension from the test preparation system, and convert the mastery status records into mastery gap markers; Read the historical review time records of the test takers in the test point dimension from the test preparation system, and convert the historical review time records into review interval markers; Combine behavioral attention markers, error cause markers, mastery gap markers, and review interval markers to create a test preparation profile record; The current preparation stage is determined based on the preparation calendar status corresponding to the recommended preparation path cycle. The target exam scope is determined based on the target exam configuration corresponding to the preparation target. The exam point dependency graph is then obtained based on the target exam scope record.

3. The test preparation path recommendation method according to claim 2, characterized in that, Based on the test preparation profile records, test point priority status identifiers are set for candidate test points in the test point dependency graph, generating a test point priority profile, including: In the test point dependency graph, candidate test points are determined based on the target test scope record; The behavioral attention markers, error cause markers, mastery gap markers, and review interval markers in the test preparation profile are mapped to candidate test points to obtain test point status mapping records corresponding to the candidate test points. Based on the test site status mapping record, generate the test site priority status identifier corresponding to the candidate test site, and generate the test site priority profile based on the candidate test site carrying the test site priority status identifier.

4. The test preparation path recommendation method according to claim 3, characterized in that, Based on the test point dependency graph, a pre-readiness check is performed on the test point priority profile. Pre-correction entries are generated for candidate test points that do not meet the pre-readiness conditions, including: Read the candidate test points carrying the test point priority status identifier from the test point priority profile, and obtain the preceding test point chain corresponding to the candidate test point from the test point dependency graph; Based on the test preparation profile record, determine the mastery status of each pre-test point in the pre-test point chain, and perform pre-test point readiness verification on the pre-test point chain according to the preset pre-test readiness conditions. If there is a candidate test point in the current test point chain that does not meet the prerequisite readiness conditions, set a downgrade flag for the candidate test point and identify the candidate test point that does not meet the prerequisite readiness conditions as a prerequisite reinforcement item. The candidate test point is generated based on a combination of the downgrade identifier and the preceding reinforcement items.

5. The test preparation path recommendation method according to claim 4, characterized in that, Preliminary reinforcement items participate in the generation of the preparation path sequence, including: After generating the prerequisite reinforcement items, determine whether there are upstream prerequisite test points in the prerequisite reinforcement items; If it exists, the corresponding upstream prerequisite test point chain is obtained according to the test point dependency graph, and the mastery status of each upstream prerequisite test point in the upstream prerequisite test point chain is determined according to the test preparation profile record. Based on the prerequisite readiness conditions, perform prerequisite readiness checks on the upstream prerequisite test point chain, and update the prerequisite reinforcement entries according to the check results. Set the pre-readiness verification completed pre-compliance entries to an insertable state, so that pre-compliance entries carrying the insertable state are arranged before the corresponding candidate test points during stage adaptation.

6. The test preparation path recommendation method according to claim 1, characterized in that, Based on the current preparation stage records, the priority profiles of test points and pre-exam correction items are adapted to the current stage to generate a sequence of preparation paths for each stage, including: Extract the phase task boundaries, test point coverage strategies, and path length control identifiers from the current preparation phase records; Based on the phase task boundaries, candidate test points for each phase are selected from the test point priority profile; The test site priority status identifier is obtained by adjusting the test site priority status identifier of the candidate test sites in the stage according to the test site coverage strategy. Based on the path length control identifier, a portion of the candidate test points carrying the stage priority status identifier are extracted, and a stage preparation path sequence is generated according to the order in which the pre-correction reinforcement entries in the pre-correction entries are located before the corresponding candidate test points, and the demotion identifiers in the pre-correction entries are applied to the corresponding candidate test points.

7. A test preparation path recommendation device, characterized in that, include: The periodic data acquisition module is used to acquire the preparation profile record, current preparation stage record, target exam scope record, and test point dependency graph corresponding to the target exam scope record when the recommended preparation path period is reached. The priority profile generation module is used to set the priority status identifier of candidate test points in the test point dependency graph based on the test preparation profile records, and generate test point priority profiles. The pre-correction generation module is used to perform pre-readiness verification on the priority profile of test points based on the test point dependency graph, and generate pre-correction entries for candidate test points that do not meet the pre-readiness conditions. The stage path generation module is used to adapt the test point priority profile and the pre-correction items to the current test preparation stage records, and generate a stage test preparation path sequence. The question type matching and generation module is used to obtain the question semantic records according to the stage preparation path sequence, semantically align the question semantic records with the test point semantic records, and generate question type matching records. The recommendation feedback update module is used to generate a preparation path recommendation record based on the stage preparation path sequence and question type matching record, and update the preparation profile record based on the execution feedback of the preparation path recommendation record.

8. An electronic device, characterized in that, It includes a processor and a memory, the memory storing computer instructions, and the processor for reading and executing the computer instructions to implement the alternative path recommendation method of any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, A computer-readable storage medium stores computer instructions that, when executed by a processor, implement the alternative path recommendation method of any one of claims 1 to 6.

10. A test preparation path recommendation system, characterized in that, This includes front-end test preparation learning devices and a cloud-based test preparation path recommendation platform; The front-end exam preparation learning device is used to: collect exam preparation process data from the exam candidates, send the exam preparation process data to the cloud-based exam preparation path recommendation platform, and display the exam preparation path recommendation records returned by the cloud-based exam preparation path recommendation platform; the front-end exam preparation learning device is also used to collect path execution feedback generated after the exam candidates execute the exam preparation path recommendation records, and send the path execution feedback to the cloud-based exam preparation path recommendation platform; The cloud-based exam preparation path recommendation platform is used to: maintain exam preparation profile records based on exam preparation process data; maintain current exam preparation stage records based on exam preparation path recommendation cycles; and maintain exam point dependency graphs based on target exam scope records. It also generates exam point priority profiles based on exam preparation profile records and exam point dependency graphs, performs pre-readiness checks on exam point priority profiles based on exam point dependency graphs, generates pre-correction items, and performs stage adaptation of exam point priority profiles and pre-correction items based on current exam preparation stage records to generate stage-specific exam preparation path sequences. Semantic alignment of question semantic records and test point semantic records is performed based on the phased preparation path sequence to generate question type matching records, and preparation path recommendation records are generated based on the phased preparation path sequence and question type matching records. The test preparation profile record is updated based on the feedback from the path execution, so that the updated test preparation profile record can participate in the next test preparation path recommendation cycle.

Images