Online paperless language teaching interactive methods, media and equipment

By constructing students' personalized basic database and language algorithm models, a multi-dimensional question combination is generated, which solves the problems of repeated mistakes in online teaching, improves learning efficiency and understanding ability, and reduces learning fatigue.

CN118656485BActive Publication Date: 2024-11-12UNION COLLEGE OF FUJIAN NORMAL UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411144324.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-11-12
Estimated Expiration
2044-08-20

AI Technical Summary

Technical Problem

The recurrence of wrong questions in existing online teaching leads to students' memory rather than understanding, reduces learning efficiency, increases learning fatigue, and deviates from the meaning of wrong questions training.

Method used

By obtaining students' learning level and learning objective information, building a basic database, filtering appropriate subject vocabulary and grammar, using language algorithm models to generate multi-dimensional question combinations, dynamically adjusting the difficulty and content of questions, and generating error-prone vocabulary and grammar group, gradually improve students' understanding of right and wrong questions.

Benefits of technology

It has achieved teaching according to the aptitude, enhanced the multi-dimensional divergence of wrong questions training, maintained the dynamic and diversification of question combinations, improved learning efficiency, and reduced students' learning fatigue.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118656485B_ABST
    Figure CN118656485B_ABST
Patent Text Reader

Abstract

The present invention relates to an online paperless language teaching interaction method, medium and device. A basic database is constructed for each student, and a current stage learning database is obtained according to the basic database, so as to realize teaching students in accordance with their aptitudes for students with different learning levels. A vocabulary group and a grammar group that are easy to make mistakes are constructed according to wrong questions, so as to realize multi-dimensional divergence of right and wrong questions. A first question group is extracted from the vocabulary group that are easy to make mistakes, and a second question group is extracted from the grammar group that are easy to make mistakes, and the first question group is inserted into the question combination of the next round. While strengthening the training of wrong questions in the next round, the wrong questions are not only focused on, but the question combination of the next round can also include questions of other difficulty levels, and the difficulty distribution of the question combination of each round at the student end is taken into consideration, so that the generation of the question combination of each round is dynamic and diversified, which is helpful to improve the learning efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of language teaching, and in particular to an online paperless language teaching interactive method, medium and device. Background Art

[0002] With the development of Internet technology, online teaching has become the main direction for updating the existing teaching model. In actual application, online teaching includes many aspects such as controlling the progress of classroom learning, teaching students in accordance with their learning levels, and automatically generating test papers and grading. In the process of test paper composition, repeated training of easy-to-make mistakes generated by students is involved. In the existing technical solutions related to wrong question training, most of them rely on summarizing students' wrong questions and then increasing the probability of repeated appearance of wrong questions to deepen students' impression of wrong questions. However, this technical solution has the following problems: the repeated appearance of wrong questions in a short period of time will cause students to remember the wrong questions instead of understanding them. At the same time, the degree of learning is concentrated on the training of the same easy-to-make mistakes and difficult questions, which is easy to increase students' learning fatigue, but deviates from the meaning of wrong question training and reduces learning efficiency. Summary of the invention

[0003] In view of the above problems, the present invention provides an online paperless language teaching interactive method, medium and device, which solves the problem that the existing wrong question screening mechanism is too rigid and reduces learning efficiency.

[0004] To achieve the above objectives, in a first aspect, the present invention provides an online paperless language teaching interactive method, comprising:

[0005] S1. Obtain the first user information of the student terminal, and construct a basic database corresponding to the current user according to the first user information, wherein the first user information includes the learning level information, grade information, grade information and subject information of the current user, and the basic database contains subject vocabulary, subject grammar content and subject question type corresponding to the current subject information;

[0006] S2, obtaining second user information, the second user information including the current user's stage learning goal and mastery level, and selecting subject vocabulary, subject grammar content and subject question types that are compatible with the stage learning goal in the basic database according to the second user information to obtain a stage learning database;

[0007] S3. Extract subject vocabulary and subject grammar content from the stage learning database and input them into the language algorithm model. The output result of the language algorithm model is a plurality of basic questions. The combination of questions of the current round corresponding to the learning objectives of the current stage is formed according to the plurality of basic questions, and the combination of questions of the current round is displayed on the student end;

[0008] S4. After the student terminal finishes answering, the actual wrong question information and the actual wrong question rate corresponding to the question combination of the current round are obtained, the actual wrong question information includes multiple wrong questions, and a plurality of first question groups and second question groups are generated according to the actual wrong question information, and the plurality of first question groups and second question groups are inserted into the question combination of the next round, and the question combination of the next round is displayed on the student terminal;

[0009] S5, generating an acceptance error rate threshold corresponding to the current user's stage learning goal according to the second user information, judging whether the actual error rate is less than the acceptance error rate threshold, if so, sending the actual error information and the actual error rate to the teacher end, if not, repeating step S4 until the actual error rate is less than the acceptance error rate threshold;

[0010] Step S4 also includes:

[0011] Follow these steps for each wrong question:

[0012] Obtain the subject vocabulary and subject grammar contained in the wrong questions, and record them as easy-to-make mistakes vocabulary and easy-to-make mistakes grammar;

[0013] Generate an error-prone vocabulary group according to the error-prone vocabulary, the error-prone vocabulary group includes the current error-prone vocabulary and confusing vocabulary, the confusing vocabulary includes vocabulary with similar semantics, similar affixes and similar word forms to the error-prone vocabulary, randomly extract a first preset number of vocabulary from the error-prone vocabulary group, and input the extracted first preset number of vocabulary into the language algorithm model, the output result of the language algorithm model is a plurality of first error-prone questions, and the plurality of first error-prone questions form a first question group;

[0014] A fallible grammar group is generated according to the fallible grammar, the fallible grammar group includes phrases corresponding to the current fallible grammar and confusing phrases, the confusing phrases include phrases with similar semantics and usage to the phrases corresponding to the current fallible grammar, a second preset number of phrases are randomly selected from the fallible grammar group, and the second preset number of phrases are input into the language algorithm model, the output result of the language algorithm model is a plurality of second fallible questions, and the plurality of second fallible questions are combined into a second question group.

[0015] Furthermore, step S4 also includes:

[0016] Obtaining the first repetition error rate of each wrong question in the actual wrong question information of the current round, where the first repetition error rate is the error rate of the question combination of the easy-to-make vocabulary group corresponding to each wrong question in the actual wrong question information in the historical rounds within a preset period;

[0017] Determine whether the first repetition error rate is less than the first error threshold, if so, delete the easy-to-make-error vocabulary group corresponding to the current wrong question, and stop inserting the first question group corresponding to the current wrong question into the question combination of the next round;

[0018] If not, generating a first adjustment weight value corresponding to the first preset number according to the first repetition error rate, and adjusting the first preset number according to the first adjustment weight value;

[0019] And, step S4 also includes:

[0020] Obtain the second repetition error rate of each wrong question in the actual wrong question information of the current round, where the second repetition error rate is the error rate of the question combination of the easy-to-make grammar group corresponding to each wrong question in the actual wrong question information in the historical rounds within a preset period;

[0021] Determine whether the second repetition error rate is less than the second error threshold, if so, delete the error-prone grammar group corresponding to the current wrong question, and stop inserting the second question group corresponding to the current wrong question into the question combination of the next round;

[0022] If not, a second adjustment weight value of a second preset number is generated according to the second repetition error rate, and the second preset number is adjusted according to the second adjustment weight value.

[0023] Furthermore, randomly selecting a first preset number of words from the error-prone vocabulary group, and generating a first question group based on the first preset number of words also includes:

[0024] A first extracted list and a first unextracted list are set, and after extracting a first preset number of words from the error-prone word group, the first preset number of words are moved from the first unextracted list to the first extracted list;

[0025] After the next round of question combinations are generated, actual wrong question information of the next round is obtained, wrong questions in the current first question group are screened out from the actual wrong question information of the next round, and vocabulary corresponding to the wrong questions in the current first question group is moved from the first extracted list to the first unextracted list;

[0026] Randomly selecting a second preset number of phrases from the error-prone grammar group, and generating a second question group based on the second preset number of phrases further includes:

[0027] Setting a second extracted list and a second unextracted list, after extracting a second preset number of phrases from the error-prone grammar group, moving the second preset number of phrases from the second unextracted list into the second extracted list;

[0028] After the next round of question combinations are generated, the actual wrong question information of the next round is obtained, the wrong questions in the current second question group are screened out from the actual wrong question information of the next round, and the phrases corresponding to the wrong questions in the current second question group are moved from the second extracted list to the second unextracted list.

[0029] Furthermore, step S4 further includes:

[0030] Obtain multiple learning knowledge points that the student has answered in the question combination of the current round, and generate the predicted mastery rate corresponding to each learning knowledge point in the current round one by one according to the actual wrong question rate;

[0031] Generate a target mastery rate based on the current user's learning level information;

[0032] The predicted mastery rates are judged one by one according to the target mastery rates. If the predicted mastery rate of a certain learning knowledge point is lower than the target mastery rate, a temporary database is constructed according to the current learning knowledge point. The temporary database includes the vocabulary, grammar and question types involved in the current learning knowledge point.

[0033] A third question group is generated according to the temporary database, and the third question group is inserted into the question combination of the next round.

[0034] Furthermore, the topic combination for the next round is obtained through the following steps:

[0035] Obtaining difficulty configuration information, the difficulty configuration information including the combination ratio of questions and question types of different difficulty levels in the question combination of the next round, the difficulty level including any one of difficult, relatively difficult, easy, and relatively easy;

[0036] The difficulty of the first and second question groups is obtained based on the repeated error rate of wrong questions;

[0037] The difficulty level of the third question group is obtained based on the predicted mastery rate of the learning knowledge points;

[0038] According to the difficulty configuration information, the first question group, the second question group and the third question group are combined and proportioned according to the preset question-taking cycle to obtain a question combination for the next round.

[0039] Furthermore, the second user information also includes consolidation target information, and the method further includes:

[0040] Obtain the historical stage learning objectives corresponding to the consolidation target information in the historical records;

[0041] Obtain the current user's historical topic combination based on the historical stage learning objectives;

[0042] A number of questions are randomly selected from the historical question combination according to the current user's mastery level to form a fourth question group;

[0043] Insert the fourth question set into the question combination of the current round.

[0044] Furthermore, the method also includes:

[0045] Building a real exam question database based on the first user information, the real exam question database includes real exam questions of previous years corresponding to the current grade information and subject information;

[0046] Obtain the predicted mastery rate of the current user. If the predicted mastery rate is higher than the target mastery rate, extract the real questions corresponding to the current learning knowledge point from the real question database to form the fifth question group.

[0047] Insert the fifth question set into the question combination for the next round.

[0048] Furthermore, step S3 also includes:

[0049] Perform cluster analysis on the subject vocabulary corresponding to the current subject information according to the degree of difficulty, and obtain multiple vocabulary clusters of different degrees of difficulty;

[0050] Acquire preset difficulty ratio information, and extract a number of vocabulary information in each vocabulary cluster from the current multiple vocabulary clusters of different difficulty levels according to the preset difficulty ratio information to obtain an extracted vocabulary set;

[0051] Input the extracted vocabulary set into the language algorithm model to obtain multiple vocabulary-based questions;

[0052] Establish a vocabulary extraction count table, and increase the number of extraction times of some vocabulary information extracted in the current stage by 1 in the current subject vocabulary;

[0053] and / or, performing cluster analysis on the subject grammar corresponding to the current subject information according to the degree of difficulty, to obtain a plurality of grammar clusters of different degrees of difficulty;

[0054] Acquire preset difficulty ratio information, and extract a number of grammatical information in each grammatical cluster from a plurality of grammatical clusters of different difficulty levels according to the preset difficulty ratio information to obtain an extracted grammar set;

[0055] Input the extracted grammar set into the language algorithm model to obtain multiple grammar-based questions;

[0056] A grammar extraction count table is established, and the number of extractions of some grammatical information extracted at the current stage is incremented by 1 in the current subject grammar.

[0057] In a second aspect, the present invention further provides a computer-readable storage medium storing computer program instructions, which implement the method described in the first aspect when executed by a processor.

[0058] In a third aspect, the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method described in the first aspect.

[0059] Different from the prior art, in the above technical solution, the first user information of the student end is obtained, and the basic database corresponding to the current user is constructed according to the first user information, and the second user information is obtained, the second user information includes the stage learning goal and mastery level of the current user, and the stage learning database is obtained by screening in the basic database according to the second user information, and a plurality of basic questions are generated in the stage learning database using the language algorithm model to form a question combination of the current round and displayed on the student end, after the student end finishes answering, the actual wrong question information and the actual wrong question rate corresponding to the question combination of the current round are obtained, and the first question group and the second question group are generated according to the actual wrong question information, wherein, the easy-to-make-mistake vocabulary group and the easy-to-make-mistake grammar group are generated according to the wrong questions in the actual wrong question information, the first question group is extracted from the easy-to-make-mistake vocabulary group, and the second question group is extracted from the easy-to-make-mistake grammar group, and a plurality of the first question group and the second question group are inserted into the question combination of the next round, and the question combination of the next round is displayed on the student end to perform multiple rounds of question training, and it is judged whether the actual wrong question rate is less than the acceptance wrong question rate threshold, and if so, the actual wrong question information and the actual wrong question rate are sent to the teacher end. In this technical solution, a basic database is constructed for each student, and then the current stage learning database is obtained based on the basic database, thereby achieving teaching according to the aptitude of students with different learning levels. Secondly, easy-to-make mistakes vocabulary groups and easy-to-make mistakes grammar groups are constructed based on wrong questions, thereby achieving multi-dimensional divergence of right and wrong questions. In addition, a first question group is extracted from the easy-to-make mistakes vocabulary group, and a second question group is extracted from the easy-to-make mistakes grammar group, and they are inserted into the question combination of the next round. While strengthening the training of wrong questions in the next round, it does not only focus on wrong questions. The question combination of the next round can also include questions of other difficulty levels, taking into account the difficulty distribution of the student end in the training process of each round of question combinations, making the generation of each round of question combinations dynamic and diversified, maintaining the activeness of human brain thinking in the training process of each round of question combinations, and improving learning efficiency.

[0060] The above-mentioned records related to the content of the invention are only an overview of the technical solution of the present invention. In order to enable ordinary technicians in the field to more clearly understand the technical solution of the present invention, and then implement it according to the text of the specification and the contents recorded in the drawings, and in order to make the above-mentioned purpose and other purposes, features and advantages of the present invention easier to understand, the following is a description in combination with the specific implementation mode of the present invention and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The drawings are only used to illustrate the principles, implementations, applications, features and effects of the specific embodiments of the present invention and other related contents, and shall not be considered as limiting the present invention.

[0062] In the drawings of the specification:

[0063] Figure 1It is a flowchart of S1 to S5 of the online teaching interaction method described in a specific embodiment of the present invention;

[0064] Figure 2 It is a flowchart of S4111 to S4114 in the fourth step of the online teaching interaction method described in a specific embodiment of the present invention;

[0065] Figure 3 It is a flowchart of S4121 to S4124 in the fourth step of the online teaching interaction method described in a specific embodiment of the present invention;

[0066] Figure 4 It is a flowchart of S421 to S424 in the fourth step of the online teaching interaction method described in a specific embodiment of the present invention;

[0067] Figure 5 This is a flowchart of S431 to S434 in the fourth step of the online teaching interaction method described in a specific implementation manner of the present invention. DETAILED DESCRIPTION

[0068] In order to explain in detail the possible application scenarios, technical principles, specific schemes that can be implemented, and the purposes and effects of the present invention, the following is a detailed description in combination with the specific embodiments listed and the accompanying drawings. The embodiments described herein are only used to more clearly illustrate the technical solutions of the present invention, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present invention.

[0069] Reference to "embodiment" herein means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The word "embodiment" appearing in various places in the specification does not necessarily refer to the same embodiment, nor does it particularly limit its independence or association with other embodiments. In principle, in the present invention, as long as there is no technical contradiction or conflict, the various technical features mentioned in each embodiment can be combined in any way to form a corresponding implementable technical solution.

[0070] Unless otherwise defined, the technical terms used herein have the same meanings as those generally understood by those skilled in the art to which the present invention belongs; the use of relevant terms herein is only for describing specific embodiments and is not intended to limit the present invention.

[0071] In the description of the present invention, the term "and / or" is an expression used to describe the logical relationship between objects, indicating that three relationships may exist, for example, A and / or B, which means: A exists, B exists, and A and B exist at the same time. In addition, the character " / " in this article generally indicates that the objects before and after are in a logical relationship of "or".

[0072] In the present invention, terms such as "first" and "second" are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship of quantity, priority or sequence between these entities or operations.

[0073] Without further restrictions, in the present invention, the words "include", "comprises", "has" or other similar open-ended expressions used in sentences are intended to cover non-exclusive inclusion. These expressions do not exclude the presence of additional elements in the process, method or product including the elements, so that the process, method or product including a series of elements may include not only those limited elements, but also other elements not explicitly listed, or also include elements inherent to such process, method or product.

[0074] Similar to the understanding in the Examination Guidelines, in the present invention, expressions such as "greater than", "less than", "exceed" and the like are understood to exclude the number itself; expressions such as "above", "below", "within" and the like are understood to include the number itself. In addition, in the description of the embodiments of the present invention, "multiple" means more than two (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups", "multiple times", etc., unless otherwise clearly and specifically limited.

[0075] See also Figure 1 In a first aspect, this embodiment provides an online paperless language teaching interaction method, comprising:

[0076] S1. Obtain the first user information of the student terminal, and construct a basic database corresponding to the current user according to the first user information, wherein the first user information includes the learning level information, grade information, grade information and subject information of the current user, and the basic database contains subject vocabulary, subject grammar content and subject question type corresponding to the current subject information;

[0077] S2, obtaining second user information, the second user information including the current user's stage learning goal and mastery level, and selecting subject vocabulary, subject grammar content and subject question types that are compatible with the stage learning goal in the basic database according to the second user information to obtain a stage learning database;

[0078] S3. Extract subject vocabulary and subject grammar content from the stage learning database and input them into the language algorithm model. The output result of the language algorithm model is a plurality of basic questions. The combination of questions of the current round corresponding to the learning objectives of the current stage is formed according to the plurality of basic questions, and the combination of questions of the current round is displayed on the student end;

[0079] S4. After the student terminal finishes answering, the actual wrong question information and the actual wrong question rate corresponding to the question combination of the current round are obtained, the actual wrong question information includes multiple wrong questions, and a plurality of first question groups and second question groups are generated according to the actual wrong question information, and the plurality of first question groups and second question groups are inserted into the question combination of the next round, and the question combination of the next round is displayed on the student terminal;

[0080] S5, generating an acceptance error rate threshold corresponding to the current user's stage learning goal according to the second user information, judging whether the actual error rate is less than the acceptance error rate threshold, if so, sending the actual error information and the actual error rate to the teacher end, if not, repeating step S4 until the actual error rate is less than the acceptance error rate threshold;

[0081] Step S4 also includes:

[0082] Follow these steps for each wrong question:

[0083] Obtain the subject vocabulary and subject grammar contained in the wrong questions, and record them as easy-to-make mistakes vocabulary and easy-to-make mistakes grammar;

[0084] Generate an error-prone vocabulary group according to the error-prone vocabulary, the error-prone vocabulary group includes the current error-prone vocabulary and confusing vocabulary, the confusing vocabulary includes vocabulary with similar semantics, similar affixes and similar word forms to the error-prone vocabulary, randomly extract a first preset number of vocabulary from the error-prone vocabulary group, and input the extracted first preset number of vocabulary into the language algorithm model, the output result of the language algorithm model is a plurality of first error-prone questions, and the plurality of first error-prone questions form a first question group;

[0085] A fallible grammar group is generated according to the fallible grammar, the fallible grammar group includes phrases corresponding to the current fallible grammar and confusing phrases, the confusing phrases include phrases with similar semantics and usage to the phrases corresponding to the current fallible grammar, a second preset number of phrases are randomly selected from the fallible grammar group, and the second preset number of phrases are input into the language algorithm model, the output result of the language algorithm model is a plurality of second fallible questions, and the plurality of second fallible questions are combined into a second question group.

[0086] In step S1, the first user information can be understood as the information of the student using the current student terminal. The specific acquisition method can be: in the database of the existing teaching system, the student's learning data is sorted according to the student's student number or the student's ID number in a preset format to obtain the first user information. When using the method steps shown in this embodiment, the student's ID number or student number can be used as the account identification number for storing the first user information. When the student logs in using the account identification number on the student terminal, the first user information is obtained. For ease of expression, the student currently using the student terminal is recorded as a user.

[0087] Specifically, the first user information includes the learning level information, grade information, and subject information of the current user, wherein the learning level information may include the learning ability and the degree of learning mastered by the current user, wherein the learning ability may be based on the average time required to master a knowledge point as an indicator, and this specific value may be obtained through big data calculation, such as obtaining the number of knowledge points that the current user has mastered in the current subject of the current grade and the cumulative learning time in the current subject, and dividing the cumulative learning time with the number of knowledge points that have been mastered, and the average time required to master a knowledge point may be obtained. For another example, the learning ability may also be based on the ranking data of the current user in the current grade information as an indicator, and the ranking data may be the ranking information of the average time required to master a knowledge point in the current grade, or the average ranking information of the current user in multiple exams in the current grade information, and this embodiment does not limit this. The degree of learning may be the proportion of the number of knowledge points that have been mastered in the total number of knowledge points in the current grade and current subject. Furthermore, the performance information may be the performance information of the examinations and tests taken by the current user in the current grade and current subject; the grade information may be divided according to the type of the current school, for example, if the current school is a university, the grade information may be one of freshman, sophomore, junior, senior, first-year graduate student, and second-year graduate student; the subject information may be pre-allocated according to the school, for example, the subject information corresponding to high school science students may be Chinese, mathematics, English, biology, chemistry, and physics.

[0088] In this embodiment, the basic database contains all subject vocabulary, subject grammar content and subject question types corresponding to the current subject information. For example, when the subject is English, the subject vocabulary is the vocabulary that needs to be mastered in the current grade. For example, the subject vocabulary of College English is the CET-4 English vocabulary library and the CET-6 English vocabulary library, and the subject grammar content is the grammatical application information corresponding to the vocabulary in the CET-4 English vocabulary library, or the subject grammar content is the grammatical application information corresponding to the vocabulary in the CET-6 English vocabulary library. Specifically, for example, when the grammatical application form of vocabulary A has subject application and object application, this application method can be integrated into information and mapped with vocabulary A for storage, which is the grammatical application information of vocabulary A; for another example, the use method of vocabulary B is to form a phrase with a certain preposition, and the predicate is used on the basis of the phrase, then this phrase and the application method can be organized into information, and then this information is mapped with vocabulary B for storage, which is the grammatical application information of vocabulary B; the subject question types can include question types under the English category: cloze, English reading, English writing, multiple-choice questions, fill-in-the-blank questions, listening questions, etc. It should also be noted that the construction of the basic database can be further screened based on a large database, or it can be generated in real time, such as by cleaning and screening the data on the Internet and the teaching system based on the first user information. For example, a full teaching category database can be pre-constructed to collect all the subject information and grade information teaching data currently required in the full teaching category database. For different users, different first user information can be used to construct them separately to obtain a basic database associated with each user. This method can improve the efficiency of building the basic database.

[0089] In step S2, the second user information can be obtained synchronously through the student end when obtaining the first user information, and the second user information can also be obtained through the teacher end, such as the teacher end can pre-enter the second user information corresponding to all student ends, and when the communication connection is established between the student end and the teacher end, the IP address of the student end is used to realize the distribution of the second user information. Specifically, the second user information includes the stage learning goal and mastery degree of the current user. Among them, the stage learning goal can be obtained from the learning plan input by the teacher end, or it can be automatically generated according to the chapter classification in the textbook used by the current subject. For example, the textbook used in the English subject includes the second unit of the first chapter, and the second unit has the first to fourth subsections. Then, the stage learning goal can be the learning content of the first subsection and the second subsection and the knowledge points to be mastered, or the stage learning goal can also be the entire second unit, which can be set according to actual needs. The mastery degree can be the chapters of the current subject that the user of the student end has learned so far, and the mastery degree can also be associated with the stage learning goal, that is, the chapters that the user has mastered in the current stage learning goal. This division method is suitable for the application scenario when the stage learning goal needs to be dispersed for multiple class hours for learning. According to the second user information, subject vocabulary, subject grammar content and subject question types that are compatible with the stage learning objectives are screened out in the basic database to obtain a stage learning database.

[0090] In step S3, the language algorithm model can be specifically constructed by adopting a basic neural network model. The specific training process of the neural network model can be: adopting a basic neural network model constructed by a recurrent neural network to establish a language sampling database. The language sampling database can be existing language subject books, textbooks, papers or online learning resources. The natural language processing is performed in the language sampling database. The specific steps may be: the sample text in the language sampling database is segmented according to words. The segmentation basis may be based on rule segmentation or statistical feature segmentation to obtain multiple sample words to be annotated. The sample words to be annotated are annotated according to the part of speech, including nouns, verbs, adjectives, etc. Each segmented sample sentence is identified and analyzed according to the syntactic relationship between words, and sample keywords are extracted on this basis. The sample keywords may include sample key words, sample inspection words, sample easy-to-mistake words, sample common words, etc. Further, a sample word form relationship network is established for each sample keyword. The sample word form relationship network includes words with similar word forms and words with similar meanings of each sample keyword, and a sample application relationship network is established for each sample keyword. The sample application relationship network includes the part of speech of each sample keyword and syntactic application information in the sample sentence. At the same time, a sample question type generation template is established. The sample question type generation template includes question type generation requirements for multiple-choice questions, listening questions, reading questions, and cloze questions. The sample question type generation template, multiple sample keywords in the language sampling database, the sample word form relationship network of each sample keyword, and the sample application relationship network of each sample keyword are input into the basic neural network model, and training and learning from vocabulary to question type are performed to obtain a trained neural network model, which is recorded as a language algorithm model. It should be noted that the language algorithm model described below refers to the current language algorithm model, and the current language algorithm model can be used to implement the generation steps from vocabulary to question type.

[0091] In this embodiment, subject vocabulary and subject grammar content are extracted from the stage learning database and input into the language algorithm model, wherein the subject vocabulary can be used as keywords. Specifically, step S3 may also include: preprocessing the subject vocabulary in the stage learning database with keywords to obtain the word form relationship network and application relationship network corresponding to each subject vocabulary, and at the same time, organizing the subject grammar content into the application relationship network of the subject vocabulary, obtaining a preset question type generation template, and inputting the preset question type generation template, keywords, and the word form relationship network and application relationship network corresponding to the keywords into the language algorithm model to obtain multiple basic questions.

[0092] Furthermore, a combination of questions for the current round corresponding to the current stage learning goal is formed based on multiple basic questions, and it is displayed on the student side. It should be noted that basic questions refer to multiple questions generated by the language algorithm model that meet the preset question type generation template. Basic questions can include one or more of multiple-choice questions, listening questions, fill-in-the-blank questions, reading questions, and cloze questions. The specific number of basic questions can be set according to actual needs, such as the number of each question type specified in the preset question type generation template. In this embodiment, students have multiple rounds of question training in the current stage learning goals. For the convenience of expression, several questions are randomly selected from the basic questions to form a combination of questions for the current round. The number of questions in the question combination can be set according to actual needs, and the number of rounds of the question combination can also be set to an upper limit value according to actual needs to avoid students still doing question training after the training class for the current stage learning goals is over.

[0093] In step S4, the student terminal answers the question combination of the current round. After the answer is completed, the answer result of the question combination of the current round is screened to obtain the actual wrong question information and the actual wrong question rate. The actual wrong question rate can be obtained by counting the number of wrong questions in the question combination of the current round, and dividing the number of wrong questions by the total number of questions in the question combination of the current round. The actual wrong question information includes the information of each wrong question. Specifically, the information of each wrong question includes the type of wrong question, the examination vocabulary and examination grammar covered by the wrong question, the reasoning ideas of the wrong question, etc., and a number of first question groups and second question groups are generated according to the actual wrong question information, and a number of first question groups and second question groups are inserted into the question combination of the next round, and the question combination of the next round is displayed on the student terminal. It should be noted that the number of the first question group and the second question group is determined according to the number of wrong questions, and each wrong question can generate a first question group and a second question group. The question combination of the next round can include questions extracted from the basic questions, and can also include some questions in the currently inserted first question group and the second question group.

[0094] Specifically, in step S4, the proportion of basic questions in the question combination of each round can be obtained in advance. In the question combination of the first round, all the questions in the question combination are extracted from the basic questions. In the question combination of the next round, the basic question adjustment proportion of the basic questions can be set. When the actual wrong question rate of the user in the question combination of the first round is high, it will lead to a large number of first question groups and second question groups in the next round. The proportion of the first question group and the second question group is set as the wrong question adjustment proportion, and the wrong question adjustment proportion and the basic question adjustment proportion can be reflected using a linear relationship. For example, the least squares method can be used for dynamic adjustment to make the first question group and the second question group have a large number of wrong questions. The number of questions in the second question group and the basic questions in the question combination of the next round can be automatically adjusted; for example, the nonlinear least squares method can be set to dynamically adjust the adjustment ratio of wrong questions and the adjustment ratio of basic questions; for example, the attention feedback parameter of each user can be introduced. The attention feedback parameter reflects the attention time and impression memory of the user when the error is made on this wrong question. By using the attention feedback parameter, the proportion of the first question group and the second question group corresponding to this wrong question in each round can be set in multiple rounds of question combination training, so as to help users deepen the impression of the current wrong question, from instantaneous memory and short-term memory to long-term memory, and realize the user's wrong question correction.

[0095] Specifically, the attention feedback parameter can be obtained in the following way: record the user's historical rounds of wrong question training in the historical stage learning goals. For example, the user has completed a total of 10 rounds of question combinations in stage learning goal A. Wrong question A appeared in the first round. In the second round, the number of the first question group and the second question group derived from wrong question A were 3 and 3 respectively. In the third round, the number of the first question group and the second question group derived from wrong question A were 1 and 1 respectively. However, in the fifth round, the number of the first question group and the second question group derived from wrong question A were 4 and 4 respectively. At this time, the error rate of the first question group and the second question group corresponding to the wrong question A of the user in the second round was relatively low, which also reduced the number of questions in the first question group and the second question group of wrong question A in the third round. However, in the fifth round of training, the user's error rate increased again, indicating that the user's short-term attention span lasts from 3 to 4 rounds. In this case, the training of wrong question A can be increased in the fifth round. After multiple statistics of wrong question data, the user's answers to the question combinations in the historical rounds of the historical stage learning goals can be refined to obtain the user's short-term attention duration and long-term attention duration for wrong questions. The short-term attention duration and long-term attention duration are integrated to obtain attention feedback parameters. For example, setting the attention feedback parameter to 0-10, for wrong question A, in the first round, the user's attention feedback parameter is 5, indicating that the attention to the wrong question is in a medium state. At this time, the user will consciously pay attention to the cause of the wrong question and analyze the wrong question; in the second round, the attention feedback parameter is 3, indicating that the user will focus on the questions similar to the current wrong question A in combination with the questions because he paid more attention to the wrong question A in the previous round, which belongs to the test memory of the exam, and effectively reduces the error rate of wrong question A; in the third round, because the error rate of wrong question A is greatly reduced, the user's attention is focused on the next wrong question, and the attention feedback parameter is 1. By analogy, when the attention feedback parameter reaches 0, the training of questions of the same category or attribute as wrong question A can be re-added to further deepen the user's impression of wrong questions in the category of wrong question A, and so on. It should be noted that the attention feedback parameter is highly subjective and needs to be obtained after data analysis based on each user's historical data. Therefore, each person's attention feedback parameter is different, which is more in line with each user's learning ability.

[0096] In this embodiment, step S4 also includes a specific method for generating the first question group and the second question group. The steps described above are performed on the wrong questions in the actual wrong question information:

[0097] Obtain the subject vocabulary and subject grammar contained in the wrong question, and record them as easy-to-make mistakes vocabulary and easy-to-make mistakes grammar. Specifically, the wrong question can be divided into word units to obtain multiple division words, the semantics of each division word can be matched, and the part of speech of each division word can be marked, the syntactic relationship of the division word in the current wrong question can be identified, and the syntactic information of each division word can be obtained, and the grammatical relationship involved in the syntactic information can be sorted out to obtain the subject grammar, and the semantics and part of speech of the division words in the wrong question can be regularized, and the vocabulary tested by the current wrong question can be obtained by combining the syntactic information identification, and recorded as subject vocabulary. It should be noted that each wrong question may have the examination of subject vocabulary and subject grammar, each wrong question may only have the examination of subject vocabulary, or each wrong question may only have the examination of subject grammar. Specifically, in the process of sorting out the grammatical relationship of the current wrong question, the identified grammatical relationship can be paired with the subject grammar in the current stage learning goal. If it does not belong to the subject grammar in the current stage learning goal, it means that the grammatical relationship contained in this wrong question does not belong to the examination scope, that is, the current wrong question does not contain subject grammar. Similarly, if there is no vocabulary matching the subject vocabulary in the current stage learning goal during the semantic and part-of-speech regularization of the current wrong question, it means that the current wrong question does not contain subject vocabulary. In the actual application process, the wrong question must contain one of the subject vocabulary and subject grammar. If the current wrong question does not contain subject vocabulary or subject grammar, it means that the current wrong question is abnormal, and the abnormal question can be sent to the teacher for further wrong question identification.

[0098] After obtaining the easy-to-make mistakes vocabulary, an easy-to-make mistakes vocabulary group is generated based on the easy-to-make mistakes vocabulary. The specific generation process can be: introduce the SIM similarity algorithm model, input the easy-to-make mistakes vocabulary into the SIM similarity algorithm model, and obtain vocabulary with similar semantics, similar affixes, and similar word forms to the easy-to-make mistakes vocabulary, match the output results of the SIM similarity algorithm model with the vocabulary in the user's basic database, and after deleting the vocabulary that is not in the current basic database, obtain the confusing vocabulary. That is, the confusing vocabulary all comes from the basic database, which can increase the user's training on the vocabulary required to master in the basic database. Optionally, when the output results of the SIM similarity algorithm model contain too few words, it means that the current easy-to-make mistakes vocabulary belongs to uncommon words, then the output results of the SIM similarity algorithm model can be not screened to avoid confusing vocabulary being 0.

[0099] A first preset number of words are randomly extracted from the easy-to-make-mistake vocabulary group, and the extracted first preset number of words are input into the language algorithm model to obtain multiple first easy-to-make-mistake questions, and the multiple first easy-to-make-mistake questions are sorted into a combination, that is, the first question group. It should be noted that the first preset number can be preset. The number of first easy-to-make-mistake questions output by the language algorithm model can also be set according to actual needs. Specifically, the number of confusing words and easy-to-make-mistake words contained in each first easy-to-make-mistake question can also be specifically set, so that the frequency of occurrence of the first preset number of words in multiple first easy-to-make-mistake questions is relatively balanced.

[0100] Similarly, after obtaining the error-prone grammar, an error-prone grammar group can be generated according to the error-prone grammar. It should be noted that the application of grammar involves the collocation between words. For the convenience of expression, this embodiment embodies the error-prone grammar in the form of phrases. In addition to the application of some fixed phrase collocations, the grammatical application of some adjectives can also be embodied in the form of phrases with part of speech collocations. Under this premise, the specific generation process of the error-prone grammar group can be: introduce the SIM similarity algorithm model, input the error-prone grammar into the SIM similarity algorithm model, and obtain a phrase with similar semantics and application methods to the phrase of the error-prone grammar, match the output result of the SIM similarity algorithm model with the subject grammar content in the user's basic database, and delete the subject grammar content that is not in the current basic database to obtain a confusing phrase. That is, the confusing phrases all come from the basic database, which can increase the user's training on the subject grammar content and phrases required to be mastered in the basic database. Optionally, when the output result of the SIM similarity algorithm model contains too few phrases, it means that the phrase corresponding to the current error-prone grammar is an uncommon phrase, and the output result of the SIM similarity algorithm model may not be filtered to avoid confusing phrases being zero.

[0101] A second preset number of phrases are randomly extracted from the error-prone grammar group, and the extracted second preset number of phrases are input into the language algorithm model to obtain multiple second error-prone questions, and the multiple second error-prone questions are organized into a combination, that is, a second question group. It should be noted that the second preset number can be preset. The number of second error-prone questions output by the language algorithm model can also be set according to actual needs. Specifically, the number of confusing phrases and phrases corresponding to the error-prone grammar contained in each second error-prone question can also be specifically set, so that the frequency of occurrence of the second preset number of phrases in multiple second error-prone questions is relatively balanced.

[0102] It should be noted that one wrong question may correspond to a first question group and a second question group. In the process of extracting question combinations for the next round, some of the first easy-to-make mistakes questions in the first question group may be selected, as may some of the second easy-to-make mistakes questions in the second question group.

[0103] In some optional embodiments, in order to avoid the user having too many wrong questions, resulting in too many first and second question groups, the following screening steps can be added to the actual wrong question information:

[0104] Determine the proportion of the number of wrong questions in the actual wrong question information to the current question combination. If it exceeds a certain proportion, for example, exceeds 80%, it means that the current user has a low level of mastery of basic questions;

[0105] The wrong questions in the actual wrong question information are clustered according to the problem-solving ideas, and wrong question clusters with the same problem-solving ideas are obtained. Then, for each wrong question cluster, the steps of generating easy-to-make mistakes vocabulary groups and easy-to-make mistakes grammar groups are executed. This method can reduce the calculation operations of the first question group and the second question group for each wrong question, and improve the generation time of the next round of question combinations.

[0106] In step S5, the acceptance error rate threshold corresponding to the current user's stage learning goal is generated according to the second user information, that is, the acceptance error rate threshold is used to indicate that the user has fully mastered the knowledge points to be tested at the current stage and can carry out learning and training for the next stage learning goal. If the actual error rate is less than the acceptance error rate threshold, it means that the user has been able to master the knowledge points to be tested at the current stage. If the actual error rate is greater than the acceptance error rate threshold, it means that the user has not fully mastered the knowledge points to be tested at the current stage, and step S4 needs to be repeated so that the user can perform multiple rounds of question combination training until the actual error rate is less than the acceptance error rate threshold.

[0107] This embodiment constructs a basic database for each student, and then obtains the current stage learning database based on the basic database, thereby achieving teaching students in accordance with their aptitude for different learning levels. Secondly, an easy-to-make mistakes vocabulary group and an easy-to-make mistakes grammar group are constructed according to wrong questions, thereby achieving multi-dimensional divergence of right and wrong questions. In addition, a first question group is extracted from the easy-to-make mistakes vocabulary group, and a second question group is extracted from the easy-to-make mistakes grammar group, and they are inserted into the question combination of the next round. While strengthening the training of wrong questions in the next round, it does not only focus on wrong questions. The question combination of the next round can also include questions of other difficulty levels, taking into account the difficulty distribution of the student end in the training process of each round of question combinations, making the generation of each round of question combinations dynamic and diversified, maintaining the activeness of human brain thinking in the training process of each round of question combinations, and improving learning efficiency.

[0108] See also Figure 2 and Figure 3 In some embodiments, step S4 further includes:

[0109] S4111, obtaining the first repetition error rate of each wrong question in the actual wrong question information of the current round, where the first repetition error rate is the error rate of the question combination of the easy-to-make vocabulary group corresponding to each wrong question in the actual wrong question information in the historical rounds within a preset period;

[0110] S4112, determining whether the first repetition error rate is less than the first error threshold;

[0111] S4113: If yes, the easy-to-make-mistake vocabulary group corresponding to the current wrong question is deleted, and the first question group corresponding to the current wrong question is stopped from being inserted into the question combination of the next round;

[0112] S4114: If not, generating a first adjustment weight value corresponding to the first preset number according to the first repetition error rate, and adjusting the first preset number according to the first adjustment weight value;

[0113] And, step S4 also includes:

[0114] S4121, obtaining the second repetition error rate of each wrong question in the actual wrong question information of the current round, where the second repetition error rate is the error rate of the question combination of the easy-to-make grammar group corresponding to each wrong question in the actual wrong question information in the historical rounds within a preset period;

[0115] S4122, determining whether the second repetition error rate is less than a second error threshold;

[0116] S4123, if yes, then delete the easy-to-make-mistake grammar group corresponding to the current wrong question, and stop inserting the second question group corresponding to the current wrong question into the question combination of the next round;

[0117] S4124: If not, generate a second adjustment weight value of a second preset number according to the second repetition error rate, and adjust the second preset number according to the second adjustment weight value.

[0118] In this embodiment, the judgment of the first repetition error rate and the second repetition error rate of the wrong questions in the actual wrong question information is introduced. It should be noted that the first repetition error rate corresponds to the error rate of the first question group generated by the vocabulary in the easy-to-make vocabulary group, and the second repetition error rate corresponds to the error rate of the second question group generated by the phrase in the easy-to-make grammar group. The application methods of the first repetition error rate and the second repetition error rate are relatively similar. This embodiment specifically describes the application method of the first repetition error rate, and the application method of the second repetition error rate can be understood by referring to the application method of the first repetition error rate.

[0119] In step S4111, the first repetition error rate of each wrong question in the actual wrong question information of the current round is obtained. It should be noted that each round of question combination will generate an actual wrong question information after answering. In this embodiment, starting from the answer record of the first round of question combination, a first repetition error rate is generated for each wrong question. The preset period is the period required for the user to complete the learning goal of the current stage, and the preset period contains training of multiple rounds of question combinations. The first repetition error rate is the cumulative error rate during the training process of multiple rounds of question combinations (that is, historical rounds of question combinations) that have been answered in the preset period for the current wrong question. For example, if the preset period has generated 10 rounds of question combination training, during the user's training of these 10 rounds of question combinations, the questions of the first question group derived from wrong question A have errors in 6 of the 10 rounds, and the first repetition error rate is 60%; it should be noted that the proportion of the first repetition error rate is changing dynamically, for example, in the 11th round of question combinations, the questions of the first question group derived from wrong question A still have 6 rounds of errors in 11 rounds, and the first repetition error rate becomes 54% after the user completes the 11th round of question combinations. In other words, the first repetition error rate of the user is gradually reduced in multiple rounds of training.

[0120] In step S4112, it is determined whether the first repetition error rate is less than a first error threshold. In this embodiment, the first error threshold can be inferred based on the learning level of the current user.

[0121] In step S4113, if the first repetition error rate is less than the first error threshold, it means that the user has mastered the knowledge point corresponding to the current wrong question, then the easy-to-make mistakes vocabulary group of the current wrong question will be deleted, and the insertion of the first question group corresponding to the current wrong question in the next round of question combinations will be stopped. This method can realize the dynamic adjustment of the same or similar wrong questions, reduce the probability of repeated appearance of questions in the first question group of the same category in the question combination of each round, and also reduce the computing power load required in the generation process of the next round of question combinations, thereby improving the generation efficiency of question combinations.

[0122] In step S4114, if the first repetition error rate is greater than the first error threshold, indicating that the knowledge point corresponding to the current wrong question has not been mastered, a first adjustment weight value is generated according to the first repetition error rate. The first adjustment weight value can adjust the specific numerical value of the first preset number to increase the number of first easy-to-make mistakes questions in the first question group, thereby increasing the number of first easy-to-make mistakes questions in the first question group in the next round of question combinations, so as to increase the number of times the user trains on questions in the same category to which the current wrong question belongs.

[0123] In some embodiments, randomly selecting a first preset number of words from the error-prone vocabulary group, and generating a first question group based on the first preset number of words further includes:

[0124] A first extracted list and a first unextracted list are set, and after extracting a first preset number of words from the error-prone word group, the first preset number of words are moved from the first unextracted list to the first extracted list;

[0125] After the next round of question combinations are generated, actual wrong question information of the next round is obtained, wrong questions in the current first question group are screened out from the actual wrong question information of the next round, and vocabulary corresponding to the wrong questions in the current first question group is moved from the first extracted list to the first unextracted list;

[0126] Randomly selecting a second preset number of phrases from the error-prone grammar group, and generating a second question group based on the second preset number of phrases further includes:

[0127] Setting a second extracted list and a second unextracted list, after extracting a second preset number of phrases from the error-prone grammar group, moving the second preset number of phrases from the second unextracted list into the second extracted list;

[0128] After the next round of question combinations are generated, the actual wrong question information of the next round is obtained, the wrong questions in the current second question group are screened out from the actual wrong question information of the next round, and the phrases corresponding to the wrong questions in the current second question group are moved from the second extracted list to the second unextracted list.

[0129] It should be noted that each round of question combination will obtain an actual wrong question information, and the actual wrong question information contains multiple wrong questions. The actual wrong question information of each round can be integrated, such as, the first easy-to-make mistakes question obtained from the actual wrong question information of the previous round is wrong again in the question combination of the current round, then this first easy-to-make mistakes question is recorded in the actual wrong question information of the question combination of the current round, and the easy-to-make mistakes vocabulary group of the previous round to which this first easy-to-make mistakes question belongs can be retained and used first, so as to reduce the generation steps of easy-to-make mistakes vocabulary group and easy-to-make mistakes grammar group in the current actual wrong question information. Optionally, even if it is the same easy-to-make mistakes vocabulary group, the first preset number of words extracted in the question combination of each round is different, and the first easy-to-make mistakes corresponding to the first preset number of words is different. That is, if the easy-to-make mistakes vocabulary group A has errors in the question combination of three rounds, then in the question combination of three rounds, the easy-to-make mistakes vocabulary group A can extend three first question groups A, the first easy-to-make mistakes in each first question group A is different, and the first preset number of words used in each first question group A is different.

[0130] In this application scenario, in this embodiment, the number of words extracted from the error-prone word group is also tracked and recorded, for example, a first extracted list and a first unextracted list are set up for the error-prone word group obtained in the current round of question combinations, and the extracted first preset number of words are moved from the first unextracted list to the first extracted list. The next round of question combinations are generated for the words in the first extracted list.

[0131] At the same time, actual wrong question information of the next round is obtained. If the actual wrong question information of the next round still contains wrong questions in the current first question group, the vocabulary corresponding to the wrong questions in the current first question group is moved from the first extracted list to the first unextracted list, and the new first preset number of vocabulary continues to be screened in the updated first unextracted list to generate a new first question group, and so on.

[0132] This method can increase the probability that the word will be selected in the next question combination when the user gives an incorrect answer to the same word, but it does not only conduct single and repetitive training on the same word. Instead, it selectively combines the same word with other words to generate new first-line wrong-answer questions for answering, so that the repetitive training of the same word is limited to a more reasonable range, and the entire wrong-answer training process will not be rigid and has a certain degree of flexibility.

[0133] Similarly, the process of generating the second question set for the error-prone grammar group is similar to the process of generating the first question set for the error-prone vocabulary group, which can be understood by referring to the above text, and this embodiment will not be described in detail.

[0134] See also Figure 4 In some embodiments, step S4 further includes:

[0135] S421, obtaining multiple learning knowledge points that the student has answered in the question combination of the current round, and generating the predicted mastery rate corresponding to each learning knowledge point in the current round one by one according to the actual wrong question rate;

[0136] S422, generating a target mastery rate according to the current user's learning level information;

[0137] S423, judging the predicted mastery rates one by one according to the target mastery rates, if the predicted mastery rate of a certain learning knowledge point is lower than the target mastery rate, constructing a temporary database according to the current learning knowledge point, the temporary database including the vocabulary, grammar and question types involved in the current learning knowledge point;

[0138] S424. Generate a third question group according to the temporary database, and insert the third question group into the question combination of the next round.

[0139] In step S421, the generation of the predicted mastery rate may include the following steps: dividing the current question combination according to multiple learning knowledge points to obtain a question set corresponding to each learning knowledge point, obtaining the distribution information of the wrong questions corresponding to the actual wrong question rate in the multiple question sets, obtaining the wrong question probability for each learning knowledge point, performing difference processing on the wrong question probability to obtain the actual mastery rate of each learning knowledge point, and using this actual mastery rate as the predicted mastery rate.

[0140] It should be noted that in step S422, the target mastery rate, that is, the mastery rate when the user fully masters a single learning knowledge point, and the mastery rates of multiple learning knowledge points are consistent. Because the user's learning level information is in a stable rising range, the target mastery rate can be obtained through the user's learning level information.

[0141] In step S423, the predicted mastery rates corresponding to multiple learning knowledge points are judged one by one. If the predicted mastery rate of a certain learning knowledge point is lower than the target mastery rate, a temporary database is constructed based on the current learning knowledge point. The temporary database includes the vocabulary, grammar and question types involved in the current learning knowledge point.

[0142] In step S424, a third question group is generated based on the temporary database. Specifically, the data in the temporary database can be input into the language algorithm model to obtain multiple temporary training questions. The multiple temporary training questions are sorted into the third question group, that is, the third question group is inserted into the question combination of the next round.

[0143] This embodiment can realize the intensive training of a single learning knowledge point. Compared with the training of the first and second question groups based on the wrong questions mentioned above, this method can realize the division of a certain knowledge point in the stage learning goal. At the same time, it enriches the question categories in the question combination in the next round, reduces the obvious wrong question emphasis concepts in the question combination, and reduces the user's sensitivity to wrong questions.

[0144] See also Figure 5 In some embodiments, the combination of questions for the next round is obtained by the following steps:

[0145] S431, obtaining difficulty configuration information, the difficulty configuration information including the combination ratio of questions and question types of different difficulty levels in the question combination of the next round, the difficulty level including any one of difficult, relatively difficult, easy, and relatively easy;

[0146] S432, obtaining the difficulty of the first question group and the second question group according to the repetition error rate of the wrong questions;

[0147] S433, obtaining the difficulty level of the third question group according to the predicted mastery rate of the learning knowledge points;

[0148] S434. According to the difficulty configuration information, the first question group, the second question group, and the third question group are combined and matched according to a preset question-solving cycle to obtain a question combination for the next round.

[0149] In step S431, the difficulty configuration information can be obtained through the teacher side. The difficulty configuration information is a combination ratio of the difficulty of the question obtained by the teacher side according to the answer situation of the user on the current student side. The difficulty level can include any one of difficult, relatively difficult, easy, and relatively easy. Optionally, the difficulty level can also be reflected as a percentage, such as 10%, 30%, 50%... and so on.

[0150] In step S432, the difficulty of the first question group and the second question group is obtained according to the repetition error rate of the wrong questions, wherein the repetition error rate includes the first repetition error rate and the second repetition error rate. It should be noted that when the repetition error rate of the user is high, it means that the current wrong question is a difficult question for the user, that is, different users have different difficulty levels in the same first question group and the second question group.

[0151] In step S433, the difficulty level of the third question set may be obtained based on the predicted mastery rate of the learning knowledge points.

[0152] In step S434, the first question group, the second question group and the third question group are combined and matched according to the difficulty configuration information according to the preset question-taking cycle to obtain the question combination for the next round. The preset question-taking cycle is also the preset cycle mentioned above.

[0153] It is understandable that the difficulty configuration information given by the teacher end will result in different contents of the next round of question combinations in different student ends. The reason is that users of different student ends have different judgment criteria for the difficulty of questions. The method shown in this embodiment can not only achieve the effect of teaching students in accordance with their aptitude in the next round of question combinations between different student ends, but also avoid plagiarism and discussion between adjacent student ends during the question-solving process, thereby improving learning efficiency and maintaining classroom discipline.

[0154] In some embodiments, the second user information further includes consolidation target information, and the method further includes:

[0155] Obtain the historical stage learning objectives corresponding to the consolidation target information in the historical records;

[0156] Obtain the current user's historical topic combination based on the historical stage learning objectives;

[0157] A number of questions are randomly selected from the historical question combination according to the current user's mastery level to form a fourth question group;

[0158] Insert the fourth question set into the question combination of the current round.

[0159] In this embodiment, the consolidation target information can be obtained through the teacher terminal, that is, the teacher can input the important knowledge points that need to be consolidated in the current learning stage on the teacher terminal. It can be understood that the consolidation target information is the learning target of the stage that the user has learned.

[0160] The historical stage learning target corresponding to the consolidation target is obtained in the historical records, and the historical question combination of the current user is obtained in the learning cycle corresponding to the historical stage learning target.

[0161] It should be noted that the current user's mastery level is the user's mastery level when learning this consolidation learning goal, and a number of questions are randomly selected from the historical question combination to form the fourth question group.

[0162] Insert the fourth question set into the question combination of the current round.

[0163] This method can increase the diversity of questions in the current round of question combinations. At the same time, it integrates the knowledge points that the current user needs to learn with the knowledge points that the user has already learned. It can be connected to the questions corresponding to the consolidation learning goals that have been learned but still need to be consolidated, completing a closed-loop training of knowledge points from learning to consolidation, which is conducive to improving the user's learning efficiency.

[0164] In some embodiments, the method further comprises:

[0165] Building a real exam question database based on the first user information, the real exam question database includes real exam questions of previous years corresponding to the current grade information and subject information;

[0166] Obtain the predicted mastery rate of the current user. If the predicted mastery rate is higher than the target mastery rate, extract the real questions corresponding to the current learning knowledge point from the real question database to form the fifth question group.

[0167] Insert the fifth question set into the question combination for the next round.

[0168] In this embodiment, a real question database is constructed according to the first user information. It should be noted that the real question database may include real question data of previous years' exams. For example, when the user's grade information is a freshman in college, the English real question data that can be obtained may be the test questions of the freshman in college in previous years and the test questions of the College English Level 4. When the user's grade information is a junior in high school, the English real question data that can be obtained may be the real questions of the college entrance examination in previous years, as well as the real questions of the English final exams and midterm exams of colleges and universities. The construction method of the real question database is similar to the construction method of the basic database, and this embodiment will not be repeated here.

[0169] The predicted mastery rate of the current user is obtained. If the predicted mastery rate is higher than the target mastery rate, it means that the user has mastered the current knowledge point very well and can practice with real questions. Then, the real questions corresponding to the current learning knowledge point are extracted from the real question database to form the fifth question group, and the fifth question group is inserted into the question combination of the next round.

[0170] This embodiment is suitable for users who have fully mastered the stage learning objectives. They can further improve their learning ability based on the fifth question set, so that users with different learning levels can get question training at corresponding levels, thereby improving learning efficiency.

[0171] In some embodiments, step S3 further includes:

[0172] Perform cluster analysis on the subject vocabulary corresponding to the current subject information according to the degree of difficulty, and obtain multiple vocabulary clusters of different degrees of difficulty;

[0173] Acquire preset difficulty ratio information, and extract a number of vocabulary information in each vocabulary cluster from the current multiple vocabulary clusters of different difficulty levels according to the preset difficulty ratio information to obtain an extracted vocabulary set;

[0174] Input the extracted vocabulary set into the language algorithm model to obtain multiple vocabulary-based questions;

[0175] Establish a vocabulary extraction count table, and increase the number of extraction times of some vocabulary information extracted in the current stage by 1 in the current subject vocabulary;

[0176] and / or, performing cluster analysis on the subject grammar corresponding to the current subject information according to the degree of difficulty, to obtain a plurality of grammar clusters of different degrees of difficulty;

[0177] Acquire preset difficulty ratio information, and extract a number of grammatical information in each grammatical cluster from a plurality of grammatical clusters of different difficulty levels according to the preset difficulty ratio information to obtain an extracted grammar set;

[0178] Input the extracted grammar set into the language algorithm model to obtain multiple grammar-based questions;

[0179] A grammar extraction count table is established, and the number of extractions of some grammatical information extracted at the current stage is incremented by 1 in the current subject grammar.

[0180] This embodiment shows the specific process of constructing basic questions. Specifically, basic questions include vocabulary basic questions and grammar basic questions. The generation process of the two is similar. This embodiment uses vocabulary basic questions for specific explanation. It should be noted that the difficulty level shown in this embodiment is an assessment of the difficulty level of subject vocabulary under normal conditions, and does not include the assessment factor of the subjective difficulty level of individual users. That is, this difficulty level allocation is based on the difficulty level of normal subject vocabulary. The difficulty level of the basic questions obtained does not have the assessment factor of the subjective difficulty level of individuals, which is equivalent to the difficulty level in the first round of question combinations in the stage learning goal not being affected and adjusted by the current user's error rate.

[0181] The subject vocabulary corresponding to the current subject information is clustered according to the degree of difficulty to obtain a plurality of vocabulary clusters of different degrees of difficulty; it should be noted that the cluster analysis can use K-means clustering, and the specific K value can be set according to actual needs. The preset difficulty ratio information is obtained, and the preset difficulty ratio information can be obtained by inputting the teacher end. In some optional embodiments, the preset difficulty ratio information and the difficulty configuration information described above can be the same information, that is, if the teacher end does not modify the preset difficulty ratio information, then this preset difficulty ratio information can be directly used to generate multiple subsequent rounds of question combinations.

[0182] It should be noted that the preset difficulty ratio information is determined according to the number of vocabulary clusters. For example, if the number of vocabulary clusters is 5, the preset difficulty ratio information also includes the difficulty ratio of the 5 vocabulary clusters. According to the preset difficulty ratio information, a number of vocabulary information in each vocabulary cluster is extracted from the current multiple vocabulary clusters of different difficulty levels to obtain an extracted vocabulary set. Specifically, the vocabulary extraction in the vocabulary cluster can be realized in the form of simple random sampling. Finally, the extracted vocabulary set is input into the language algorithm model to obtain a vocabulary-based question.

[0183] This embodiment sets preset difficulty ratio information to achieve a reasonable combination of questions of multiple difficulty levels in the process of generating basic questions. The configuration of basic questions matches the user's question-solving thinking. The combination of basic questions from shallow to deep can better mobilize the user's desire to learn and improve the user's learning efficiency.

[0184] In a second aspect, this embodiment further provides a computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described in the first aspect is implemented.

[0185] In a third aspect, this embodiment further provides an electronic device, comprising a memory and a processor, wherein the memory is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method described in the first aspect.

[0186] The storage medium / memory includes but is not limited to: RAM, ROM, disk, tape, CD, flash memory, USB flash drive, mobile hard disk, memory card, memory stick, network server storage, network cloud storage, etc. The processor includes but is not limited to CPU (central processing unit), GPU (graphics processing unit), MCU (microprocessor), etc.

[0187] In the above technical solution, the first user information of the student end is obtained, and the basic database corresponding to the current user is constructed according to the first user information, and the second user information is obtained, the second user information includes the stage learning goal and mastery level of the current user, and the stage learning database is obtained by screening in the basic database according to the second user information, and a plurality of basic questions are generated in the stage learning database using the language algorithm model to form a question combination of the current round and displayed on the student end, after the student end finishes answering, the actual wrong question information and the actual wrong question rate corresponding to the question combination of the current round are obtained, and the first question group and the second question group are generated according to the actual wrong question information, wherein the easy-to-make-mistake vocabulary group and the easy-to-make-mistake grammar group are generated according to the wrong questions in the actual wrong question information, the first question group is extracted from the easy-to-make-mistake vocabulary group, and the second question group is extracted from the easy-to-make-mistake grammar group, and a plurality of the first question group and the second question group are inserted into the question combination of the next round, and the question combination of the next round is displayed on the student end to perform multiple rounds of question training, and it is judged whether the actual wrong question rate is less than the acceptance wrong question rate threshold, and if so, the actual wrong question information and the actual wrong question rate are sent to the teacher end. In this technical solution, a basic database is constructed for each student, and then the current stage learning database is obtained based on the basic database, thereby achieving teaching according to the aptitude of students with different learning levels. Secondly, easy-to-make mistakes vocabulary groups and easy-to-make mistakes grammar groups are constructed based on wrong questions, thereby achieving multi-dimensional divergence of right and wrong questions. In addition, a first question group is extracted from the easy-to-make mistakes vocabulary group, and a second question group is extracted from the easy-to-make mistakes grammar group, and they are inserted into the question combination of the next round. While strengthening the training of wrong questions in the next round, it does not only focus on wrong questions. The question combination of the next round can also include questions of other difficulty levels, taking into account the difficulty distribution of the student end in the training process of each round of question combinations, making the generation of each round of question combinations dynamic and diversified, maintaining the activeness of human brain thinking in the training process of each round of question combinations, and improving learning efficiency.

[0188] Finally, it should be noted that although the above embodiments have been described in the specification and drawings of the present invention, the patent protection scope of the present invention cannot be limited thereby. All technical solutions generated by replacing or modifying equivalent structures or equivalent processes based on the essential concept of the present invention and using the contents recorded in the specification and drawings of the present invention, as well as directly or indirectly implementing the technical solutions of the above embodiments in other related technical fields, are included in the patent protection scope of the present invention.

Claims

1. An online paperless language teaching interactive method, characterized in that: include: S1. Obtain first user information of the student terminal, and construct a basic database corresponding to the current user according to the first user information, wherein the first user information includes the learning level information, grade information, grade information and subject information of the current user, and the basic database contains subject vocabulary, subject grammar content and subject question type corresponding to the current subject information; S2, obtaining second user information, wherein the second user information includes the current user's stage learning goal and mastery level, and selecting subject vocabulary, subject grammar content and subject question types that are compatible with the stage learning goal from the basic database according to the second user information to obtain a stage learning database; S3, extracting subject vocabulary and subject grammar content from the stage learning database, and inputting them into the language algorithm model, the output result of the language algorithm model is a plurality of basic questions, and forming a current round of question combinations corresponding to the current stage learning objectives according to the plurality of basic questions, and displaying the current round of question combinations on the student end; S4. After the student terminal finishes answering, actual wrong question information and actual wrong question rate corresponding to the question combination of the current round are obtained, wherein the actual wrong question information includes multiple wrong questions, and a plurality of first question groups and second question groups are generated according to the actual wrong question information, and the plurality of the first question groups and the second question groups are inserted into the question combination of the next round, and the question combination of the next round is displayed on the student terminal; S5, generating an acceptance error rate threshold corresponding to the current user's stage learning goal according to the second user information, judging whether the actual error rate is less than the acceptance error rate threshold, if so, sending the actual error information and the actual error rate to the teacher end, if not, repeating step S4 until the actual error rate is less than the acceptance error rate threshold; Step S4 also includes: For each of the wrong questions, perform the following steps: Obtaining the subject vocabulary and subject grammar contained in the wrong question, and recording them as easy-to-make mistakes vocabulary and easy-to-make mistakes grammar; Generating an error-prone vocabulary group according to the error-prone vocabulary, wherein the error-prone vocabulary group includes current error-prone vocabulary and confusing vocabulary, wherein the confusing vocabulary includes vocabulary with similar semantics, similar affixes, and similar word forms to the error-prone vocabulary, randomly extracting a first preset number of vocabulary from the error-prone vocabulary group, and inputting the extracted first preset number of vocabulary into a language algorithm model, wherein the output result of the language algorithm model is a plurality of first error-prone questions, and the plurality of first error-prone questions form the first question group; Generate an error-prone grammar group according to the error-prone grammar, the error-prone grammar group includes phrases corresponding to the current error-prone grammar and confusing phrases, the confusing phrases include phrases with similar semantics and usage to the phrases corresponding to the current error-prone grammar, randomly extract a second preset number of phrases from the error-prone grammar group, and input the second preset number of phrases into the language algorithm model, the output result of the language algorithm model is a plurality of second error-prone questions, and the plurality of second error-prone questions form the second question group; The language algorithm model is obtained according to the following method: A basic neural network model constructed using a recurrent neural network was used to establish a language sampling database; Performing natural language processing in a language sampling database, specifically including: segmenting the sample text in the language sampling database according to words to obtain a plurality of sample words to be annotated, annotating the sample words to be annotated according to parts of speech, identifying and analyzing each segmented sample sentence according to the syntactic relationship between words, and extracting sample keywords on this basis, wherein the sample keywords include sample key words, sample inspection words, sample easy-to-error words, and sample commonly used words; A sample word form relationship network is established based on each sample keyword, the sample word form relationship network includes words with similar word forms and words with similar meanings to each sample keyword, and a sample application relationship network is established for each sample keyword, the sample application relationship network includes the part of speech of each sample keyword and syntactic application information in the sample sentence; A sample question generation template is set up. The sample question generation template includes question generation requirements for multiple-choice questions, listening questions, reading questions, and cloze questions. The sample question generation template, multiple sample keywords in the language sampling database, the sample word form relationship network of each sample keyword, and the sample application relationship network of each sample keyword are input into the basic neural network model to conduct training and learning from vocabulary to question types, and a trained neural network model is obtained, which is recorded as a language algorithm model.

2. The paperless online language teaching interactive method according to claim 1 is characterized in that: Step S4 also includes: Obtaining a first repetition error rate of each wrong question in actual wrong question information of the current round, wherein the first repetition error rate is an error rate of a question combination of a prone-to-mistake vocabulary group corresponding to each wrong question in the actual wrong question information in a historical round within a preset period; Determine whether the first repetition error rate is less than a first error threshold, and if so, delete the easy-to-make-error vocabulary group corresponding to the current wrong question, and stop inserting the first question group corresponding to the current wrong question into the question combination of the next round; If not, generating a first adjustment weight value corresponding to a first preset number according to the first repetition error rate, and adjusting the first preset number according to the first adjustment weight value; And, step S4 also includes: Obtaining a second repetition error rate of each wrong question in the actual wrong question information of the current round, wherein the second repetition error rate is an error rate of a combination of questions in historical rounds within a preset period for the easy-to-make grammar group corresponding to each wrong question in the actual wrong question information; Determine whether the second repetition error rate is less than a second error threshold, and if so, delete the error-prone grammar group corresponding to the current wrong question, and stop inserting the second question group corresponding to the current wrong question into the question combination of the next round; If not, generating a second adjustment weight value of a second preset number according to the second repetition error rate, and adjusting the second preset number according to the second adjustment weight value.

3. The paperless online language teaching interactive method according to claim 2 is characterized in that: Randomly selecting a first preset number of words from the error-prone vocabulary group, and generating the first question group according to the first preset number of words further includes: Setting a first extracted list and a first unextracted list, after extracting a first preset number of words from the error-prone vocabulary group, moving the first preset number of words from the first unextracted list into the first extracted list; After the next round of question combinations are generated, actual wrong question information of the next round is obtained, wrong questions in the current first question group are screened out from the actual wrong question information of the next round, and vocabulary corresponding to the wrong questions in the current first question group is moved from the first extracted list to the first unextracted list; Randomly selecting a second preset number of phrases from the error-prone grammar group, and generating the second question group according to the second preset number of phrases further includes: Setting a second extracted list and a second unextracted list, after extracting a second preset number of phrases from the fallible grammar group, moving the second preset number of phrases from the second unextracted list into the second extracted list; After the next round of question combinations are generated, the actual wrong question information of the next round is obtained, the wrong questions in the current second question group are screened out from the actual wrong question information of the next round, and the phrases corresponding to the wrong questions in the current second question group are moved from the second extracted list to the second unextracted list.

4. The paperless online language teaching interactive method according to claim 3 is characterized in that: Step S4 also includes: Acquire multiple learning knowledge points that the student has answered in the question combination of the current round, and generate a predicted mastery rate corresponding to each learning knowledge point in the current round one by one according to the actual wrong question rate; Generate a target mastery rate based on the current user's learning level information; The predicted mastering rates are judged one by one according to the target mastering rates. If the predicted mastering rate of a certain learning knowledge point is lower than the target mastering rate, a temporary database is constructed according to the current learning knowledge point, wherein the temporary database includes vocabulary, grammar and question types involved in the current learning knowledge point; A third question group is generated according to the temporary database, and the third question group is inserted into the question combination of the next round.

5. The paperless online language teaching interactive method according to claim 4 is characterized in that: The next round of question combinations are obtained through the following steps: Obtaining difficulty configuration information, wherein the difficulty configuration information includes a combination ratio of questions and question types of different difficulty levels in the question combination of the next round, wherein the difficulty level includes any one of difficult, relatively difficult, easy, and relatively easy; Obtaining the difficulty of the first question group and the second question group according to the repetition error rate of the wrong questions; Obtaining the difficulty level of the third question set according to the predicted mastery rate of the learning knowledge points; The first question group, the second question group and the third question group are combined and matched according to the difficulty configuration information according to a preset question-taking cycle to obtain a question combination for the next round.

6. The paperless online language teaching interactive method according to claim 5 is characterized in that: The second user information also includes consolidation target information, and the method further includes: Obtain the historical stage learning goal corresponding to the consolidation goal information in the historical record; Acquire a combination of historical questions of the current user according to the historical stage learning goal; A number of questions are randomly selected from the historical question combination according to the current user's mastery level to form a fourth question group; The fourth question group is inserted into the question combination of the current round.

7. The paperless online language teaching interactive method according to claim 6 is characterized in that: The method further comprises: Building a real exam question database according to the first user information, the real exam question database including real exam questions of previous years corresponding to the current grade information and subject information; Obtaining the predicted mastery rate of the current user, and if the predicted mastery rate is higher than the target mastery rate, extracting real questions corresponding to the current learning knowledge point from the real question database to form a fifth question group; The fifth question group is inserted into the question combination of the next round.

8. The paperless online language teaching interactive method according to claim 7 is characterized in that: Step S3 also includes: Perform cluster analysis on the subject vocabulary corresponding to the current subject information according to the degree of difficulty, and obtain multiple vocabulary clusters of different degrees of difficulty; Acquire preset difficulty ratio information, and extract a number of vocabulary information in each of the vocabulary clusters of multiple different difficulty levels according to the preset difficulty ratio information to obtain an extracted vocabulary set; Inputting the extracted vocabulary set into a language algorithm model to obtain a plurality of vocabulary-based questions; Establishing a vocabulary extraction count table, and increasing the number of extraction times of a number of vocabulary information extracted in the current subject vocabulary by 1; and / or, performing cluster analysis on the subject grammar corresponding to the current subject information according to the degree of difficulty, to obtain a plurality of grammar clusters of different degrees of difficulty; Acquire the preset difficulty ratio information, and extract a number of grammatical information in each of the grammatical clusters of the current multiple grammatical clusters of different difficulty levels according to the preset difficulty ratio information to obtain an extracted grammar set; Inputting the extracted grammar set into a language algorithm model to obtain a plurality of grammar-based questions; A grammar extraction count table is established, and the number of extractions of a number of grammatical information extracted at the current stage in the current subject grammar is incremented by 1.

9. A computer-readable storage medium storing computer program instructions, characterized in that: The computer program instructions implement the method according to any one of claims 1 to 8 when executed by a processor.

10. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method according to any one of claims 1 to 8.

Citation Information

Patent Citations

  • Wrong question statistical analysis system

    CN113761030A

  • Personalized intelligent test paper composition

    CN118445325A

  • Intelligent exercise recommendation method and apparatus, computer device and storage medium

    WO2021253480A1

  • Question-answering area acquisition method and system, electronic assignment processing method and system, and device and medium

    WO2024149409A1