CN108897879B

CN108897879B - Method for realizing personalized teaching through man-machine interaction

Info

Publication number: CN108897879B
Application number: CN201810725586.6A
Authority: CN
Inventors: 孙一乔
Original assignee: Beijing Xizi Education Technology Co ltd
Current assignee: Beijing Sita Intelligent Technology Co.,Ltd.
Priority date: 2018-07-04
Filing date: 2018-07-04
Publication date: 2020-11-10
Anticipated expiration: 2038-07-04
Also published as: CN108897879A

Abstract

The invention relates to the technical field of artificial intelligence and discloses a method for realizing personalized teaching through human-computer interaction. It provides a personalized teaching method for realizing teaching links including automatic decomposition of learning tasks, automatic recommendation of test questions, positioning of error steps, question explanation and the like through man-machine interaction, the method needs no real teacher to participate in other steps except the data preparation and parameter setting in the first step, thus reducing the occupation of human resources, especially education resources to the maximum extent, meanwhile, compared with the traditional explanation mode which takes 15-20 minutes on average, the method can not only furthest skip the content which does not need to be explained aiming at the specific learning condition of the user and mainly explain the doubts of the user by determining the error factors and carrying out step-level precision explanation, and the problem source is deeply excavated, so that the user can understand the knowledge points more easily, the explanation time can be shortened to 2-3 minutes, and the learning efficiency of the user is greatly improved.

Description

Method for realizing personalized teaching through man-machine interaction

Technical Field

The invention belongs to the technical field of artificial intelligence, and particularly relates to a method for realizing personalized teaching through human-computer interaction, which can be applied to the education industry.

Background

In the education industry, realizing personalized education is a great important subject. At present, the exploration of the existing online education technology in the aspect of individuation, namely adaptive learning, can only be stopped on the technologies of live broadcast, video teaching, individualized operation/question bank and the like.

The live broadcast technology is divided into 1-to-1 live broadcast, large-class live broadcast and other applications. In the 1-to-1 live broadcast, each teacher can only serve one student at the same time, namely, each student needs to be equipped with a single teacher, so that the method is extremely high in cost and difficult to popularize in consideration of the tension of teacher resources, especially excellent teacher resources. In a live broadcast of a large class, a teacher needs to consider the conditions of all students simultaneously when facing hundreds or even thousands of students, so that it is impossible to perform a personalized teaching mode for each student, such as answering and confusion for a certain student, and the teaching is not targeted (or personalized) basically.

In the video teaching technology, students learn by watching given teaching videos, and recorded videos do not have any change, and all students see the same content, so that no personalized possibility exists naturally. It is a good attempt to let different students watch different videos, but because the video recording cost is very high, it is the limit that one knowledge point records 2-3 videos, and personalizing different conditions of the students will lead to the increase of the required number of videos at an exponential speed, so increasing the number of videos to adapt to the students only can carry out very coarse personalization, and cannot really refine and give a sufficiently targeted explanation for each student.

In addition to the above two technologies, the most popular adaptive education technology is based on personalized question bank/homework. In the technology, the system evaluates students by collecting the exercise results of the students and recommends more exercise exercises according to the evaluation results, thereby realizing personalization to a certain extent from the perspective of exercise. But when the student makes mistakes, the student cannot determine the specific position of the wrong subject, which directly results in that the unit of explanation is the subject rather than a certain step in the subject, so that the student has to waste more time on the content which does not need to be learned. Meanwhile, the explanation still depends on the above-mentioned video teaching mode, so that the problem of difficult personalized explanation still cannot be solved.

Disclosure of Invention

In order to solve the above problems in the prior art, the present invention aims to provide a method for implementing personalized teaching through human-computer interaction.

The technical scheme adopted by the invention is as follows:

a method for realizing personalized teaching through human-computer interaction comprises the following steps:

s100, pre-storing basic data of knowledge points, current learning data of a user and basic data of test questions of a plurality of test questions to be answered in a database, wherein the basic data of the knowledge points comprise a plurality of learning ranges, knowledge points corresponding to the learning ranges and a first topological sequence for expressing the checking relation of all the knowledge points in sequence, the current learning data of the user comprise a knowledge point set learned by the user, a test question set done by the user and the current capability value of the user at each knowledge point, and the basic data of the test questions comprise question contents, standard answers with at least two answering steps, answer skills corresponding to at least one answering step, a second topological sequence for expressing the checking relation of all the answering steps in sequence and knowledge points, knowledge point difficulty values and knowledge point distinguishing degree values corresponding to each answering step;

s101, acquiring a target learning range and a target ability value of a user through man-machine interaction;

s102, obtaining a target learning knowledge point set corresponding to the target learning range according to the corresponding relation between the learning range and the knowledge points;

s103, adjusting the target learning knowledge point set according to the first topological order and the current learning data of the user;

s104, eliminating knowledge points of which the current ability value of the user exceeds the target ability value from the target learning knowledge point set, if elements in the target learning knowledge point set are zero, executing a step S113, otherwise, selecting the target learning knowledge point located at the most prior position in the first topological order from the target learning knowledge point set as a current prior learning knowledge point;

s105, aiming at the current prior learning knowledge point, calculating the current adaptation degree of each test question according to the knowledge point basic data, the current learning data of the user and the test question basic data of all answer test questions, and taking the test question with the highest current adaptation degree as a prior teaching test question;

s106, sending the question content of the priority teaching test question to a human-computer interaction interface, and obtaining a response result of the user through human-computer interaction;

s107, if the answer result is not consistent with the standard answer corresponding to the priority teaching test question, judging that the answer is wrong, and executing the step S108, otherwise, judging that the answer is correct, and executing the step 112;

s108, positioning a pre-estimation error answer step of the user answering this time according to the current learning data of the user and the basic data of the test questions of the prior teaching test questions;

s109, outputting and displaying the estimated error answer step and all prior answer steps which are positioned before the estimated error answer step according to the second topological order and are not marked with human-computer interaction results on a human-computer interaction interface, and then marking the human-computer interaction results of the displayed answer steps through human-computer interaction, wherein the human-computer interaction results indicate that the answer steps are marked to be question or question-free correspondingly;

s110, explaining corresponding explanation materials pre-stored in a database aiming at the display solution step newly marked as the questioning and/or aiming at the problem solving skills comprising the display solution step;

s111, in all the answering steps of the prior teaching test questions, if the answering step without marking the man-machine interaction result still exists, returning to execute the steps S108-S111;

s112, updating the current learning data of the user according to a correct answer result of the user answering the current time or the human-computer interaction record generated in the steps S109-110, and then returning to execute the steps S104-S112;

and S113, finishing the learning, and outputting learning finishing information to a human-computer interaction interface, wherein the learning finishing information comprises the current capacity value promotion amount of each knowledge point of the user in the target learning range.

Optimally, in the step S103, the target learning knowledge point set is adjusted as follows in the manner described by the steps S301 to S302 and/or in the manner described by the steps S303 to S304:

s301, aiming at each knowledge point in the target learning knowledge point set, searching all prior knowledge points which are positioned in front of the knowledge point and have intervals of which the number is not more than a preset distance value according to the first topological order;

s302, aiming at each prior knowledge point, if the current ability value of the user at the prior knowledge point is lower than the target ability value, adding the prior knowledge point into the target learning knowledge point set;

s303, aiming at each knowledge point in the target learning knowledge point set, searching all prior knowledge points which are positioned in front of the knowledge point and have the interval number larger than a preset distance value and posterior knowledge points which are positioned behind the knowledge point according to the first topological order;

s304, aiming at each posterior knowledge point, if the current ability value of the user at the posterior knowledge point is lower than the target ability value, pushing the posterior knowledge point to a human-computer interaction interface, and if the posterior knowledge point is confirmed to be learned through human-computer interaction, adding the posterior knowledge point to the target learning knowledge point set.

Optimally, in the step S105, the current adaptation degree of each test question is calculated as follows:

s501, respectively calculating the following indexes of the corresponding test questions according to the current prior learning knowledge point, the target capacity value, the basic data of the knowledge point, the current learning data of the user and the basic data of the test questions of the corresponding test questions: ratio F of number of knowledge points outside target in test question to number of knowledge points in test question₁The ratio F of the topological order span of the knowledge points in the test question to the number of the knowledge points in the database₂The ratio F of the number of the knowledge points in the test question to the number of the knowledge points in the test question is not mastered by the user₃The ratio F of the last-tested position of the topological order of the knowledge points in the test question to the number of the knowledge points in the database is not mastered by the user₄The ratio F of the weight of the answering step outside the inner target of the test question to the weight of the answering step inside the test question₅And/or the probability ratio F of the user missing the knowledge point outside the target₆；

S502, splicing all indexes obtained in the step S501 into a column vector;

s503, subtracting the ideal column vector corresponding to the most suitable test question from the column vector to obtain an error vector, wherein the ideal column vector is a 0 vector or is preset according to teaching experience;

and S504, taking the two norms of the error vectors as the current adaptation degree of the corresponding test questions.

Preferably, before the step S107, the method further includes the following steps:

s700, if the answering result acquired for the first time is inconsistent with the standard answer corresponding to the priority teaching test question, the question content of the priority teaching test question and the test point prompt are sent to a man-machine interaction interface, and the answering result of the user is acquired again through man-machine interaction, wherein the test point prompt is stored in the test question basic data of the priority teaching test question in advance.

Preferably, in the step S108, the step of estimating the error solution of the current answer of the user is positioned as follows:

s801, aiming at each answering step of the prior teaching test questions, calculating to obtain corresponding individual pair prediction probability according to the current learning data of the user and the basic data of the test questions;

s802, in all the answer steps of the unmarked human-computer interaction result, the answer step with the lowest prediction probability is independently taken as the estimated error answer step.

Further preferably, in step S801, the method for calculating the individual pair prediction probability of the solution step according to the current learning data of the user and the basic data of the test questions includes the following steps:

and searching the knowledge point, the knowledge point difficulty value and the knowledge point distinguishing degree value corresponding to the answering step from the test question basic data, searching the current capability value of the user at the knowledge point corresponding to the answering step from the current learning data of the user, inputting the current capability value, the knowledge point difficulty value and the knowledge point distinguishing degree value into an IRT mathematical model, and taking the output probability of the IRT mathematical model as the independent opposite prediction probability corresponding to the answering step.

Preferably, in the step S110, the display solution step is explained as follows:

s1101, generating first blackboard writing information, wherein the first blackboard writing information comprises all knowledge points corresponding to the display solution step;

s1102, sending the first blackboard writing information and a first explaining material for explaining the display answering step to a man-machine interaction interface for output and display, wherein the first explaining material comprises a picture file, a text file, a voice file and/or a video file, and is bound and stored in the test question basic data of the prior teaching test question together with the corresponding display answering step in advance;

s1103, if the user feedback is not understood, executing the step S1104, otherwise, ending the explanation of the display answering step;

s1104, generating a first knowledge point set to be explained, and bringing all knowledge points corresponding to the solution displaying step into the first knowledge point set to be explained;

s1105, aiming at each knowledge point to be explained in the first knowledge point set to be explained, searching all prior knowledge points before the knowledge point to be explained according to the first topological order;

s1106, aiming at each prior knowledge point, if the current ability value of the user at the prior knowledge point is lower than the target ability value, adding the prior knowledge point into the first knowledge point set to be explained;

and S1107, explaining the knowledge points to be explained in the first knowledge point set to be explained one by one according to the sequence of the first topological order from first to last.

Optimally, in the step S110, the problem solving skills are explained as follows:

s1111, generating second blackboard writing information, wherein the second blackboard writing information comprises all solution steps corresponding to the problem solving skills and all knowledge points corresponding to the solution steps;

s1112, sending the second blackboard writing information and second explanation materials for explaining the problem solving skills to a human-computer interaction interface for output and display, wherein the second explanation materials comprise picture files, text files, voice files and/or video files, and are bound with the corresponding problem solving skills in advance and stored in the test question basic data of the priority teaching test questions;

s1113, if the user feedback is not understood, executing a step S1114, otherwise, ending the explanation of the problem solving skills;

s1114, generating a second knowledge point set to be explained, and bringing all knowledge points corresponding to all the answering steps in all the answering steps corresponding to the question solving skills into the second knowledge point set to be explained;

s1115, aiming at each knowledge point to be explained in the second knowledge point set to be explained, searching all prior knowledge points before the knowledge point to be explained according to the first topological order;

s1116, aiming at each prior knowledge point, if the current ability value of the user at the prior knowledge point is lower than the target ability value, adding the prior knowledge point into the second knowledge point set to be explained;

and S1117, explaining the knowledge points to be explained in the second knowledge point set to be explained one by one according to the sequence of the first topological order from first to last.

Further optimized, the knowledge points to be explained are explained as follows:

s1121, determining the explanation difficulty of the knowledge points according to the current capability values of the users at the knowledge points to be explained;

s1122, correcting the explanation difficulty of the knowledge points upwards or downwards according to the learning times of the knowledge points to be explained of the user, wherein the learning times and the corresponding knowledge points to be explained are bound and stored in the current learning data of the user in advance;

s1123, sending a third explanation material which corresponds to the explanation difficulty of the knowledge points and is used for explaining the knowledge points to be explained to a man-machine interaction interface for output and display, and enabling the learning times of the knowledge points to be explained to be added by 1, wherein the third explanation material comprises picture files, text files, voice files and/or video files, and is bound with the corresponding knowledge points to be explained in advance and stored in the basic data of the knowledge points;

s1124, if the user feedback is not understood, executing the step S1125, otherwise ending the explanation of the knowledge point to be explained;

s1125, if other third explanation materials with explanation difficulty just lower than that of the knowledge point are stored in the database, sending the third explanation materials to a man-machine interaction interface for output and display, adding 1 to the learning times of the knowledge point to be explained, and then returning to the step S1124, otherwise, executing the step S1126;

s1126, if a priori knowledge point which is located in front of the knowledge point to be explained and has a user current capability value lower than the target capability value exists in the first topological sequence, taking the priori knowledge point as a new knowledge point to be explained with priority, or taking all the priori knowledge points located in front of the knowledge point to be explained as new knowledge points to be explained with priority;

s1127, explaining the new knowledge points to be explained according to the first topological order and the second topological order one by one according to the steps S1121-S1127.

Preferably, after the step S110, the method further includes the following steps:

s1131, after a certain display solution step is finished or the explanation of the problem solving skills including the certain display solution step is included, according to the second topological order, if all the prior solution steps before the display solution step are found to be marked as no question, the display solution step is marked as a true solution step, and then the step S1132 is executed, otherwise the subsequent explanation is continued;

s1132, if the answer times of the user on the priority teaching test questions do not exceed the preset value and the user confirms to continue answering through human-computer interaction, terminating the subsequent explanation, and then returning to execute the step S106, otherwise, continuing the subsequent explanation.

The invention has the beneficial effects that:

(1) the invention provides a personalized teaching method for realizing teaching links such as automatic decomposition of learning tasks, automatic recommendation of test questions, positioning of error steps, questioning explanation and the like through man-machine interaction, which needs to carry out data preparation and parameter setting in the first step, does not need any real teacher participation in other steps, reduces human resources, especially the occupation of education resources to the maximum extent, and compared with the traditional explanation mode which consumes 15-20 minutes on average, the method can skip the content which does not need to be explained to the maximum extent according to the specific learning condition of a user by determining the error factor and carrying out step-level precision explanation, mainly explains the questioning of the user, deeply excavates the problem source of the user, enables the user to understand knowledge points more easily, and can shorten the explanation time to 2-3 minutes, the learning efficiency of the user is greatly improved;

(2) compared with the traditional explanation and ability evaluation taking the questions as units, the method can perform fine knowledge point level evaluation on users giving the same result through more detailed step relations in the test questions, association and other labels between the test questions and the knowledge points, so that the method can obtain more accurate student ability evaluation results by using fewer test questions, and the more accurate evaluation results can make question recommendation and knowledge point explanation more targeted, thereby further realizing real personalized education.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

FIG. 1 is a flow chart of a method for realizing personalized teaching through human-computer interaction according to the present invention.

Detailed Description

The invention is further described with reference to the following figures and specific embodiments. It should be noted that the description of the embodiments is provided to help understanding of the present invention, but the present invention is not limited thereto.

The term "and/or" herein is merely an association describing an associated object, meaning that three relationships may exist, e.g., a and/or B, may mean: a exists alone, B exists alone, and A and B exist at the same time, and the term "/and" is used herein to describe another association object relationship, which means that two relationships may exist, for example, A/and B, may mean: a alone, and both a and B alone, and further, the character "/" in this document generally means that the former and latter associated objects are in an "or" relationship.

Example one

As shown in fig. 1, the method for implementing personalized teaching through human-computer interaction provided by this embodiment includes the following steps.

S100, pre-storing basic data of knowledge points, current learning data of a user and basic data of test questions of a plurality of test questions to be answered in a database, wherein the basic data of the knowledge points comprise a plurality of learning ranges, knowledge points corresponding to the learning ranges and a first topological sequence used for expressing the checking relation of all the knowledge points in sequence, the current learning data of the user comprise a knowledge point set learned by the user, a test question set done by the user and the current capability value of the user at each knowledge point, and the basic data of the test questions comprise question contents, standard answers with at least two answering steps, answer skills corresponding to at least one answering step, a second topological sequence used for expressing the checking relation of all the answering steps in sequence, and knowledge points, knowledge point difficulty values and knowledge point distinguishing values corresponding to each answering step.

In the step S100, the knowledge point precedence relationship refers to that, for the knowledge point a and the knowledge point b, if the knowledge point a must be learned first before the knowledge point b can be learned, the knowledge point a and the knowledge point b have a posteriori relationship: the knowledge points a are prior knowledge points, and the knowledge points b are posterior knowledge points. Thereby obtaining the first topological order which has a certain length and expresses all knowledge point precedence check relations. The checking relation of the answering step is that for the answering step a and the answering step b, if the answering step a must be carried out first and then the answering step b can be carried out, the answering step a and the answering step b have a posterior-to-anterior relation: the solution step a is a prior step, and the solution step b is a posterior step. The second topological order can thus be obtained which has a topological network structure and which expresses the a-priori relationships of all the solution steps. In addition, optimally, the test question basic data may further include examination point prompts, weight coefficients corresponding to each solution step, and the like.

S101, acquiring a target learning range and a target ability value of a user through man-machine interaction.

In the step S101, the specific manner of the human-computer interaction may be, but is not limited to, a question-and-answer teaching manner based on a virtual AI teacher role, that is, a target learning range and a target ability value of the user are obtained through a voice conversation or a text conversation between the virtual AI teacher and the user on the human-computer interface, where the target learning range may be, but is not limited to, a certain chapter in a textbook. In addition, the answer results of subsequent test questions and the human-computer interaction results of the specific answer steps can be obtained based on the same question-answer teaching method.

S102, obtaining a target learning knowledge point set corresponding to the target learning range according to the corresponding relation between the learning range and the knowledge points.

In the step S102, since the knowledge point basic data includes a plurality of learning ranges and knowledge points corresponding to the learning ranges, the learning task can be automatically decomposed at a knowledge point level, so that the subsequent targeted teaching can be performed one by one based on the first topological order.

S103, adjusting the target learning knowledge point set according to the first topological order and the current learning data of the user.

In step S103, in order to ensure that the user can achieve the capability improvement at each target learning knowledge point, the set of target learning knowledge points may be adjusted in the following manner described by steps S301 to S302 and/or in the manner described by steps S303 to S304, but is not limited to:

The method described in steps S301 to S302 is a method of forcibly adding knowledge points, and a priori knowledge points within a certain interval range must be learned because they are preconditions for learning knowledge points for learning the target. The method described in steps S303 to S304 is a method of optionally adding knowledge points, and whether to select a correction or not can be determined by human-computer interaction for a priori knowledge point and a posterior knowledge point which are outside a certain interval range. The preset distance value is a preset natural number not less than 2.

S104, eliminating the knowledge points of which the current ability value of the user exceeds the target ability value from the target learning knowledge point set, if the elements in the target learning knowledge point set are zero, executing the step S113, otherwise, selecting the target learning knowledge point located at the most prior position in the first topological order from the target learning knowledge point set as the current prior learning knowledge point.

In step S104, particularly, if it is found that a plurality of knowledge points need to be explained preferentially in the user' S question making process, the knowledge points can be selected sequentially from the knowledge points. In addition, if a certain number of test questions are already completed on a certain current priority learning knowledge point (even if the current priority learning knowledge point still needs teaching), the current priority learning knowledge point is recorded to be skipped, and then the current priority learning knowledge point is removed from the target learning knowledge point set needing teaching, so that the learning progress can be prevented from being blocked on a certain knowledge point.

And S105, aiming at the current prior learning knowledge point, calculating the current adaptation degree of each test question according to the knowledge point basic data, the current learning data of the user and the test question basic data of all answer test questions, and taking the test question with the highest current adaptation degree as a prior teaching test question.

In the step S105, the current adaptation degree of each test question may be calculated, but not limited to, as follows.

S501, respectively calculating the following indexes of the corresponding test questions according to the current prior learning knowledge point, the target capacity value, the basic data of the knowledge point, the current learning data of the user and the basic data of the test questions of the corresponding test questions: ratio F of number of knowledge points outside target in test question to number of knowledge points in test question₁The ratio F of the topological order span of the knowledge points in the test question to the number of the knowledge points in the database₂The ratio F of the number of the knowledge points in the test question to the number of the knowledge points in the test question is not mastered by the user₃The ratio F of the last-tested position of the topological order of the knowledge points in the test question to the number of the knowledge points in the database is not mastered by the user₄The ratio F of the weight of the answering step outside the inner target of the test question to the weight of the answering step inside the test question₅And/or the probability ratio F of the user missing the knowledge point outside the target₆。

In the step S501, the ratio F of the number of the knowledge points outside the test question inner target to the number of the knowledge points inside the test question inner target of the test question can be calculated according to the following steps₁：

S5101, summarizing to obtain a test question covering knowledge point set according to test question basic data of corresponding test questions;

s5102, calculating the ratio F of the number of the knowledge points outside the targets in the test question to the number of the knowledge points in the test question according to the following formula₁：

In the formula (I), the compound is shown in the specification,

the total number of knowledge points belonging to the test question covering knowledge point set and not belonging to the current prior learning knowledge point,

covering knowledge points of a set of knowledge points for belonging to said test questionAnd (4) total number. In the step S5101, since the question basic data includes the question content, the standard answers including at least two answer steps, and the knowledge points corresponding to each answer step, the test question covering knowledge point set of the test question can be easily obtained.

In the step S501, the ratio F of the topological order span of knowledge points in the test question to the number of knowledge points in the library of the test question can be calculated according to the following steps₂：

S5201, summarizing according to test question basic data of corresponding test questions to obtain a test question covering knowledge point set;

s5202, determining the prior knowledge point at the first position of the first topological order and the last-test knowledge point at the last position of the topological order in the test question covering knowledge point set according to the knowledge point basic data;

s5203, calculating the ratio F of the topological order span of the knowledge points in the test question to the number of the knowledge points in the database according to the following formula₂：

Wherein M is the total number of knowledge points in the database, X_MaxLIs the topology number, X, of the last-verified knowledge point_MinFIs the topology sequence number, X, of the most prior knowledge point_MaxL-X_MinFThe topological order span of the knowledge points in the test question. Because the knowledge point basic data comprises the first topological sequence used for expressing all the knowledge point precedence check relations, the prior knowledge point, the last knowledge point and the corresponding topological sequence number can be easily determined.

In the step S501, the ratio F of the number of knowledge points in the test question not grasped by the user to the number of knowledge points in the test question may be calculated according to the following steps₃：

S5301, summarizing according to test question basic data of corresponding test questions to obtain a test question coverage knowledge point set;

s5302, aiming at each knowledge point in the test question covering knowledge point set, finding a corresponding current capacity value from the current learning data of the user, and if the current capacity value is lower than the target capacity value, taking the knowledge point as a knowledge point in the test question which is not mastered by the user;

s2303, a knowledge point set in the test question which is not mastered by the user is obtained in a gathering mode, and the ratio F of the number of the knowledge points in the test question which is not mastered by the user to the number of the knowledge points in the test question is calculated according to the following formula₃：

In the formula (I), the compound is shown in the specification,

the total number of the knowledge points belonging to the knowledge point set in the test question which is not mastered by the user,

and covering the total number of the knowledge points in the knowledge point set for the test question. Because the current learning data of the user comprises the current ability values of the user at all knowledge points, whether a certain knowledge point can be mastered by the user or not can be easily judged.

In the step S501, the ratio F of the last-tested position of the topological order of the knowledge points in the test question, which is not grasped by the user of the test question, to the number of knowledge points in the library may be calculated, but not limited thereto, according to the following steps₄：

S5401, summarizing to obtain a test question covering knowledge point set according to test question basic data of corresponding test questions;

s5402, aiming at each knowledge point in the test question covering knowledge point set, searching a corresponding current capacity value from the current learning data of the user, and if the current capacity value is lower than the target capacity value, taking the knowledge point as a knowledge point in the test question which is not mastered by the user;

s5403, gathering to obtain a knowledge point set in the test question which is not mastered by the user;

s5404, determining the last-test non-mastered knowledge point at the last position of the topological order in the knowledge point set in the user non-mastered test question according to the knowledge point basic data;

s5405, calculating the ratio F of the last test position of the topological order of the knowledge points in the test question not mastered by the user to the number of the knowledge points in the library according to the following formula₄：

Wherein M is the total number of knowledge points in the database,

and the topology serial number of the last-a-priori unknown knowledge point is obtained.

In the step S501, the ratio F of the weight of the examination question inner target outer solution step to the weight of the examination question inner solution step of the examination question can be calculated according to the following steps₅：

S5501, summarizing to obtain a set of answer steps in the test questions according to basic data of the test questions corresponding to the test questions;

s5502, aiming at each answer step in the answer step set in the test questions, judging whether the answer step corresponds to the current prior learning knowledge point or not according to the basic data of the test questions corresponding to the test questions, and if not, taking the answer step as an answer step outside a target in the test questions;

s5503, a test question inner target outer answer step set is obtained in a gathering mode, and then the ratio F of the weight of the test question inner target outer answer step to the weight of the test question inner answer step is calculated according to the following formula₅：

In the formula, w_sFor solving outside the target in the test question step set

Weight coefficient of individual solution step, w_jTo be at the same timeIn the set of test question inner solution steps

The weight coefficients of the individual solution steps,

the total number of the solution steps belonging to the solution step set outside the test question inner target,

the total number of the solution steps belonging to the solution step set in the test question.

In the step S501, the ratio F of the probability that the user misses the off-target knowledge point₆Aiming at the answering steps in the test questions and not corresponding to the current prior learning knowledge points, the method predicts and calculates the pairing probability of the user in the answering steps according to the current learning data (containing the current ability value of the user at each knowledge point) of the user and the basic data of the test questions (containing the difficulty coefficient corresponding to each answering step), and finally calculates the probability that the user wrongly answers at least one of the answering steps, namely the probability that the user wrongly answers at the knowledge points outside the target by combining the probability theory.

S502, all the indexes obtained in the step S501 are spliced into a column vector.

S503, subtracting the ideal column vector corresponding to the most suitable test question from the column vector to obtain an error vector, wherein the ideal column vector is a 0 vector or is preset according to teaching experience.

After the step S503, in order to avoid the problem that repeated development is required to be performed while all cases need to be considered, the method further includes the following steps: and multiplying the error vector by a preset weight row vector to complete the adjustment of the error vector.

After the step S504, in order to consider both the new topic practice and the old topic review and avoid monotonicity of topic recommendation under the same ability, the method further includes the following steps:

according to the current learning data of the user, the user made test question set is counted, the number of times of making questions of the user corresponding to the test questions is counted, and upward correction is carried out on the current adaptation degree as shown in the following formula:

a′_i＝γ^ta_i

in formula (II), a'_iThe current fitting degree of the ith (i ═ 1,2,3, …, N) test question after correction, a_iThe current fitting degree of the ith (i is 1,2,3, …, N) test question before correction, gamma is an upward correction coefficient, and t is the number of times of questions made by the user on the test question. Therefore, repeated problem making can be avoided by adding correction of the number of times of making problems, but the possibility of recommendation is kept, and the purpose of review is achieved.

And S106, sending the question content of the priority teaching test question to a human-computer interaction interface, and acquiring a response result of the user through human-computer interaction.

In the step S106, the answering result may include, but is not limited to, several answering steps and final answers.

S107, if the answer result is not consistent with the standard answer corresponding to the priority teaching test question, judging that the answer is wrong, and executing the step S108, otherwise, judging that the answer is correct, and executing the step 112.

Before step S107, in order to give the user more than one opportunity to answer the error correction, the method further includes the following steps:

And S108, positioning the pre-estimation error answer step of the user answering this time according to the current learning data of the user and the basic data of the test questions of the prior teaching test questions.

In the step S108, the step of predicting the error solution of the user' S response may be, but is not limited to, positioning as follows:

In step S801, the method for calculating the individual pair prediction probability of the answer step according to the current learning data of the user and the basic data of the test questions includes the following steps:

The IRT mathematical model is an existing mathematical model based on an IRT Theory (Item Response Theory, also called topic Response Theory or latent trait Theory, which is a collective name of a series of psychology models) and used for analyzing test results or questionnaire survey data. Specifically, the IRT mathematical model may be, but is not limited to, a 3-parameter Normal-objective model or a 3-parameter Logistic model; or, calculating the individual pair prediction probability of the solution step according to the following formula:

p_sgr＝sigmoid(κ(v-d))

in the formula, sigmoid () is a nonlinear function applied in an IRT mathematical model, κ is a knowledge point discrimination value corresponding to an answer step, and d is a knowledge point difficulty value corresponding to the answer step.

S109, outputting and displaying the estimated error answer step and all prior answer steps which are positioned before the estimated error answer step according to the second topological order and are not marked with human-computer interaction results on a human-computer interaction interface, and then marking the human-computer interaction results of the displayed answer steps through human-computer interaction, wherein the human-computer interaction results indicate that the answer steps are marked to be displayed with questions or without questions correspondingly.

In step S109, particularly, after all the solution steps corresponding to a certain solution skill have been completely displayed, the solution skill needs to be displayed and interacted with the user. When the user proposes the feedback result, the result is recorded as the basis for positioning the error step. In addition, if the user has a problem with the steps displayed in the combined mode, the user can click and expand all the combined individual solution steps to display, and then marking the human-computer interaction results of the individual solution steps.

S110, aiming at the display solution step newly marked as the question and/or aiming at the problem solving skills containing the display solution step, corresponding explanation materials pre-stored in a database are applied for explanation.

In the step S110, the display solution step may be explained, but not limited to, as follows:

s1102, sending the first blackboard writing information and a first explaining material for explaining the display answering step to a man-machine interaction interface for output and display, wherein the first explaining material can be but is not limited to comprise a picture file, a text file, a voice file and/or a video file, and is bound with the corresponding display answering step in advance and stored in the test question basic data of the prior teaching test question;

In the step S110, the problem solving skills can be, but are not limited to, explained as follows:

In the step S1107 or the step S1117, the knowledge points to be explained may be explained, but not limited to, as follows:

After the step S110, in order to confirm the real error reason and consolidate the user' S understanding of the explanation knowledge in time, the following steps are further included:

S111, in all the answering steps of the priority teaching test questions, if the answering step without the marked human-computer interaction result still exists, returning to execute the steps S108-S111.

And S112, updating the current learning data of the user according to a correct answer result of the user answering the current time or the human-computer interaction record generated in the steps S109-110, and then returning to execute the steps S104-S112.

In the step S112, the current learning data of the user may be updated by, but not limited to, the following ways: and for the correct answering result, marking all the answering steps as the question-free answering steps, then for all the question-free answering steps, calling the current ability value of the user on the knowledge point corresponding to the answering step, and for all the question answering steps, adjusting the current ability value of the user on the knowledge point corresponding to the answering step downwards. Particularly, for the posterior solution step after the true error solution step according to the second topological order, if no continuous question is returned, the current ability value of the user on the knowledge point corresponding to the posterior solution step is not updated. In addition, all the solution steps of the prior teaching test questions and all the explanation records performed on the basis of the solution steps can be displayed to the user on a human-computer interaction interface, so that the student can consolidate the questions.

In summary, the method for realizing personalized teaching through human-computer interaction provided by the embodiment has the following technical effects:

(1) the embodiment provides a personalized teaching method for realizing teaching links including automatic decomposition of learning tasks, automatic recommendation of test questions, positioning of error steps, question explanation and the like through man-machine interaction, the method needs no real teacher to participate in other steps except the data preparation and parameter setting in the first step, thus reducing the occupation of human resources, especially education resources to the maximum extent, meanwhile, compared with the traditional explanation mode which takes 15-20 minutes on average, the method can not only furthest skip the content which does not need to be explained aiming at the specific learning condition of the user and mainly explain the doubts of the user by determining the error factors and carrying out step-level precision explanation, the problem source is deeply excavated, so that the user can understand the knowledge points more easily, the explanation time can be shortened to 2-3 minutes, and the learning efficiency of the user is greatly improved;

The present invention is not limited to the above-described alternative embodiments, and various other forms of products can be obtained by anyone in light of the present invention. The above detailed description should not be taken as limiting the scope of the invention, which is defined in the claims, and which the description is intended to be interpreted accordingly.

Claims

1. A method for realizing personalized teaching through human-computer interaction is characterized by comprising the following steps:

2. The method for realizing personalized education through human-computer interaction according to claim 1, wherein in the step S103, the set of target learning knowledge points is adjusted as follows in a manner described by the steps S301 to S302 and/or a manner described by the steps S303 to S304:

3. The method for implementing personalized education through human-computer interaction as claimed in claim 1, wherein in said step S105, the current adaptation degree of each test question is calculated as follows:

s501, respectively calculating the following indexes of the corresponding test questions according to the current prior learning knowledge point, the target capacity value, the basic data of the knowledge point, the current learning data of the user and the basic data of the test questions of the corresponding test questions: ratio F of number of knowledge points outside target in test question to number of knowledge points in test question₁The ratio F of the topological order span of the knowledge points in the test question to the number of the knowledge points in the database₂The ratio F of the number of the knowledge points in the test question to the number of the knowledge points in the test question is not mastered by the user₃The user does not master the topological order of the knowledge points in the test question and the final test position and the number of the knowledge points in the databaseRatio F of₄The ratio F of the weight of the answering step outside the inner target of the test question to the weight of the answering step inside the test question₅And/or the probability ratio F of the user missing the knowledge point outside the target₆；

S502, splicing all indexes obtained in the step S501 into a column vector;

4. The method for realizing personalized education through human-computer interaction according to claim 1, further comprising the following steps before the step S107:

5. The method for realizing personalized teaching through human-computer interaction as claimed in claim 1, wherein in said step S108, the step of estimating the error solution of the user' S response this time is positioned as follows:

6. The method as claimed in claim 5, wherein the step S801 of calculating the individual pair prediction probability of the solution step according to the current learning data of the user and the basic data of the test questions comprises the following steps:

7. The method for realizing personalized education through human-computer interaction according to claim 1, wherein in the step S110, the display-solution step is explained as follows:

8. The method for realizing personalized education through human-computer interaction according to claim 1, wherein in the step S110, the problem solving skills are explained as follows:

9. The method for realizing personalized education through human-computer interaction according to claim 7 or 8, characterized in that the knowledge points to be explained are explained as follows:

10. The method for realizing personalized education through human-computer interaction according to claim 1, further comprising the following steps after the step S110: