Teaching information processing method based on AI

Abstract

The invention relates to the technical field of information processing, in particular to an AI-based teaching information processing method. Comprising the steps of determining whether a generation parameter of an exclusive question set for a single user is qualified or not based on a score characterization value; when it is judged that the generation parameters of the exclusive question set are abnormal, the generation parameters of the exclusive question set for a single user are adjusted based on the wrong question discrete magnitude, and random knowledge points are added to serve as anchor points; or adjusting the preset selected number and the number of the obtained historical conventional test papers to corresponding values based on the data integrity, the identification abnormal data proportion and the historical test lack frequency, or deleting the hollow values of the historical conventional test papers used for determining the knowledge points corresponding to the wrong questions. The test question deployment of the latest knowledge points and the historical to-be-reviewed knowledge points is balanced, the processing efficiency for the teaching information is improved, and the portrait accuracy for the user is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information processing, and in particular to a teaching information processing method based on AI. BACKGROUND

[0002] In traditional teaching information processing, there are many deficiencies in personalized learning support for users. In the past, teaching often adopts a "one-size-fits-all" approach to provide the same learning materials and exercises for all users, without fully considering the knowledge mastery and learning characteristics of each user.

[0003] When determining the weak knowledge points of users, it usually relies on the subjective judgment and experience of teachers, lacking scientific and systematic methods. Moreover, the analysis of user performance is relatively simple, only focusing on the scores of single exams, making it difficult to comprehensively and accurately assess the learning status and progress of users.

[0004] When generating exercise sets, it cannot be customized according to the specific situation of users, resulting in users may repeat practicing knowledge points they have already mastered, while the knowledge points that really need to be strengthened are not adequately trained. In addition, when the user's performance fluctuates, it is difficult to timely and reasonably adjust the teaching strategy and the generation parameters of the exercise set, affecting the teaching effect and the learning efficiency of users.

[0005] Chinese patent application publication No. CN115984054A discloses a smart education method and system based on a big data platform, including a teacher terminal, a user terminal, and a management platform. The teacher terminal is used to obtain question publishing information. The user terminal is used to obtain question test information based on the question publishing information. The management platform is used to analyze the question test information to obtain an analysis result and generate question recommendation information based on the analysis result and a preset big data question bank. The teacher terminal is also used to obtain new question publishing information based on the analysis result and the question recommendation information. It can be seen that the above technical solution has the following problems: it does not consider balancing the latest knowledge points and historical review knowledge points for test question deployment, resulting in single test question deployment, which leads to low accuracy of user profiling. SUMMARY

[0006] Therefore, the present application provides a teaching information processing method based on AI to overcome the problem in the prior art that the test question deployment is not balanced between the latest knowledge points and historical review knowledge points, resulting in single test question deployment and low accuracy of user profiling.

[0007] To achieve the above purpose, the present application provides a teaching information processing method based on AI, comprising: determining the knowledge points corresponding to each feature character segment based on the identified feature character segments for a single user in the conventional database. arrange the knowledge points in descending order of frequency of occurrence, and select a preset number of knowledge points as anchor points; select a number of questions corresponding to the anchor points in the question database according to the anchor points, and generate a special question set for a single user; determine the score representation value based on the conventional score, wherein the conventional score is the score of the conventional test paper of the single user; determine whether the generation parameters of the data set for the single user are qualified based on the score representation value; When it is determined that the generation parameters of the data set are abnormal, adjust the generation parameters of the data set for the single user based on the wrong question dispersion quantity, including adding random knowledge points as anchor points; determine that the generation parameters of the data set are qualified, and continue to use the current parameters to complete the generation of the data set for the single user.

[0008] Further, the process of determining whether the generation parameters of the data set for the single user are qualified based on the score representation value includes: When the score representation value is less than or equal to the second preset score representation value and greater than the first preset score representation value, determine whether the generation parameters of the data set for the single user are qualified in combination with the horizontal comparison parameter; determine the horizontal comparison parameter based on the continuously monitored conventional scores of each user; When the horizontal comparison parameter is less than or equal to the preset horizontal comparison parameter, determine that the generation parameters of the data set are abnormal, and adjust the generation parameters of the data set for the single user based on the wrong question dispersion quantity.

[0009] Further, when the horizontal comparison parameter is greater than the preset horizontal comparison parameter, adjust the first preset score representation value to a corresponding value based on the horizontal comparison parameter.

[0010] Further, when the score representation value is greater than the second preset score representation value, determine that the generation parameters of the data set are abnormal, and adjust the generation parameters of the data set for the single user based on the wrong question dispersion quantity.

[0011] Further, adjusting the generation parameters of the data set for the single user based on the wrong question dispersion quantity includes; determine the wrong question dispersion quantity based on the obtained time node at which the single user initially learns the knowledge points corresponding to each anchor point; When the wrong question dispersion quantity is less than or equal to the preset wrong question dispersion quantity, adjust the generation parameters of the data set for the single user based on the data integrity of the current conventional test paper; When the wrong question dispersion quantity is greater than the preset wrong question dispersion quantity, determine the number of selected random knowledge points based on the wrong question dispersion quantity.

[0012] Further, based on the data integrity of the current regular test paper, the generation parameters of the data set for a single user are adjusted, including: Based on the data of each answer of the obtained current regular test paper, the null values in each data are identified, and the integrity of the data is determined based on the number of null values; When the integrity is greater than the preset integrity, the number of historical regular test papers in the obtained regular database for a single user is adjusted to a corresponding value based on the score representation value.

[0013] Further, when the integrity is less than or equal to the first preset integrity, the generation parameters of the data set for a single user are adjusted based on the identified abnormal data proportion, including: Based on the data of each answer of the obtained current regular test paper, the number of abnormal data is identified, and the abnormal data proportion is determined based on the number of identified abnormal data; When the identified abnormal data proportion is greater than the preset abnormal data proportion, an alarm information for test paper content identification is issued.

[0014] Further, when the identified abnormal data proportion is less than or equal to the preset abnormal data proportion, the generation parameters of the data set for a single user are adjusted based on the historical absence frequency of a single user, including: Based on the obtained number of absences of a single user within a preset historical time period, the historical absence frequency of a single user is determined.

[0015] Further, when the historical absence frequency is less than or equal to the preset historical absence frequency, the null values in the historical regular test paper used to determine the corresponding knowledge points of each feature character segment are deleted.

[0016] Further, when the historical absence frequency is greater than the preset historical absence frequency, the preset selected number is adjusted to a corresponding value based on the integrity.

[0017] Compared with the prior art, the beneficial effects of the present application are that whether the generation parameter of the exclusive question set for the single user is qualified is determined based on the score representation value, when the score representation value is less than or equal to the first preset score representation value, at this time, the score fluctuation of the single user is smaller, and is in a relatively stable state, at this time, it is determined that the generation parameter of the exclusive question set is qualified. When the score representation value is less than or equal to the second preset score representation value and greater than the first preset score representation value, at this time, the score of the user exists slight fluctuation, and further analysis is combined with the horizontal comparison parameter. Whether the generation parameter of the exclusive question set for the single user is qualified is determined combined with the horizontal comparison parameter, the horizontal comparison parameter represents the score fluctuation of the single user compared with other users in the same period, when the horizontal comparison parameter is less than or equal to the preset horizontal comparison parameter, at this time, the score of the single user has obviously decreased compared with the users in the same group, in this case, the generation parameter of the exclusive question set is adjusted. When the horizontal comparison parameter is greater than the preset horizontal comparison parameter, at this time, the scores of the users in the same period generally decrease, it is determined that the score fluctuation is generally large due to the abnormally high difficulty of the test paper, at this time, the first preset score representation value is adjusted. The regular scores of the users are continuously collected, and the variance of each regular score is calculated as the score representation value. According to the comparison result of the score representation value and the preset score representation value, whether the generation parameter of the exclusive question set needs to be adjusted is judged. The score fluctuation of the user is found in time, and the generation parameter of the exclusive question set is reasonably adjusted according to different situations, so that the effectiveness and pertinence of teaching are ensured. Whether the generation parameter of the exclusive question set for the single user is qualified is determined combined with the horizontal comparison parameter, the score situation of the users in the same group is considered, the misjudgment caused by external factors such as test paper difficulty is avoided, and whether the generation parameter of the exclusive question set needs to be adjusted is more accurately judged. The teaching effect and the learning efficiency of the user are improved, the processing efficiency of the teaching information is improved, and the accuracy of the portrait of the user is further improved.

[0018] Further, based on the data integrity of the current conventional test paper, the generation parameters of the exclusive question set for the single user are adjusted, the integrity represents the completion of the test paper, when the integrity is greater than the first preset integrity, at this time the user has completed most of the answers of the test paper, the score fluctuation at this time is determined as the weak knowledge point, at this time the number of historical conventional test papers is adjusted to more accurately determine the recent weak knowledge point of the user, when the data integrity is less than or equal to the first preset integrity, at this time there may be an abnormal fluctuation in the score due to the fact that the handwriting is too messy to accurately identify the answer, at this time the generation parameters of the exclusive question set are adjusted based on the proportion of abnormal data. According to the answering situation of the user and the test paper recognition situation, the generation parameters of the exclusive question set are flexibly adjusted, the adaptability and accuracy of teaching are improved. Based on the proportion of abnormal data, the generation parameters of the exclusive question set for the single user are adjusted, the abnormal data represents the data situation that cannot be recognized or misrecognized, when the proportion of abnormal data is greater than the preset proportion of abnormal data, there is a large amount of data that cannot be recognized in the test paper, an alarm information for the content recognition of the test paper is sent, the problem in the content recognition of the test paper is found in time, the accuracy of the data is ensured, a reliable basis is provided for subsequent teaching analysis and question set generation, the processing efficiency of the teaching information is improved, and the accuracy of the portrait of the user is further improved.

[0019] Further, when the proportion of abnormal data is less than or equal to the preset proportion of abnormal data, at this time the user's answer content can be clearly recognized by the program, under this condition, the score abnormally fluctuates due to the fact that the user has a situation of missing the exam, based on the historical missing exam frequency of the single user, the generation parameters of the exclusive question set for the single user are adjusted, when the historical missing exam frequency is less than or equal to the preset historical missing exam frequency, it is determined that the single user has a missing exam due to a sudden accident, at this time the null value is removed when determining the knowledge point corresponding to the wrong question, so as to improve the accuracy of the exclusive question set; when the historical missing exam frequency is greater than the preset historical missing exam frequency, the single user frequently misses the exam in the recent period, at this time the preset selected number is adjusted to focus on the key problems of the single user. The influence of the missing exam of the user on the learning effect is considered, the generation of the exclusive question set is more in line with the actual situation of the user, the processing efficiency of the teaching information is improved, and the accuracy of the portrait of the user is further improved. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The step flow chart of the AI-based teaching information processing method of the embodiment of the application; Figure 2 The logic determination diagram for determining whether the generation parameters of the exclusive question set for the single user are qualified based on the score representation value of the embodiment of the application; Figure 3 The logic determination diagram for determining whether the generation parameters of the exclusive question set for the single user are qualified in combination with the horizontal comparison parameter of the embodiment of the application; Figure 4 A logic decision diagram for adjusting the generation parameter of the exclusive question set for the single user based on the discrete quantity of the wrong question. DETAILED DESCRIPTION

[0021] In order to make the objects and advantages of the present application clearer, the present application will be further described below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0022] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. It should be understood by those skilled in the art that the embodiments are only used to explain the technical principles of the present application and are not used to limit the protection scope of the present application.

[0023] Please refer to Figure 1 , Figure 2 , Figure 3 and Figure 4 , which are respectively a step flowchart of an AI-based teaching information processing method according to an embodiment of the present application, a logic decision diagram for determining whether the generation parameter of the exclusive question set for the single user is qualified based on the score representation value, a logic decision diagram for determining whether the generation parameter of the exclusive question set for the single user is qualified in combination with the horizontal comparison parameter, and a logic decision diagram for adjusting the generation parameter of the exclusive question set for the single user based on the discrete quantity of the wrong question. An AI-based teaching information processing method according to an embodiment of the present application comprises: S1, determining the knowledge points corresponding to each feature character segment based on the identified feature character segments in the historical regular test papers for the single user; the feature character segment is the wrong question in the identified historical regular test paper; S2, arranging the knowledge points in descending order according to the quantity, and selecting the top pre-set selected quantity of knowledge points as anchor points according to the arrangement; S3, selecting the questions corresponding to the anchor points in the test question database according to the anchor points, and generating the exclusive question set for the single user; S4, determining the score representation value based on the regular score, wherein the regular score is the score of the regular test paper of the single user; S5, determining whether the generation parameter of the exclusive question set for the single user is qualified based on the score representation value; S6, when it is determined that the generation parameter of the exclusive question set is abnormal, adjusting the generation parameter of the exclusive question set for the single user based on the discrete quantity of the wrong question, including adding a random knowledge point as an anchor point, or adjusting the pre-set selected quantity, the number of the obtained historical regular test papers to the corresponding value based on the data integrity, the proportion of the identified abnormal data, and the historical missing frequency, or deleting the unanswered test questions in the historical regular test paper used to determine the knowledge points corresponding to each wrong question; S7, determining that the generation parameter of the exclusive question set is qualified, and continuing to use the current parameter to complete the generation of the exclusive question set for the single user.

[0024] Specifically, the conventional test paper is a test paper facing all users, and the exclusive question set is a question set for a single user. The conventional database stores each historical conventional test paper of each user, and the data set is the exclusive question set.

[0025] Specifically, each question in each test paper is pre-stored with a corresponding knowledge point.

[0026] In a single embodiment, a single historical conventional test paper is obtained, and five wrong questions are identified, and the knowledge points corresponding to the wrong questions are 6 chapters 2 sections, 6 chapters 3 sections, 7 chapters 2 sections, 7 chapters 4 sections and 7 chapters 2 sections.

[0027] Specifically, the test question database includes a plurality of questions, each of which is labeled with a corresponding knowledge point.

[0028] Specifically, in the S4, according to the determined anchor point knowledge points, a plurality of questions corresponding to these knowledge points are selected from the test question database to generate an exclusive question set for a single user.

[0029] Specifically, the score representation value is determined based on the conventional score; The variance of a plurality of conventional scores is calculated to obtain a score representation value for a single user; The process of determining whether the generation parameter of the exclusive question set for the single user is qualified based on the score representation value includes: If the score representation value is less than or equal to the first preset score representation value, it is determined that the generation parameter of the exclusive question set is qualified, and the current parameter is continued to be used to complete the generation of the exclusive question set for the single user; If the score representation value is less than or equal to the second preset score representation value and greater than the first preset score representation value, it is determined whether the generation parameter of the exclusive question set for the single user is qualified in combination with the transverse comparison parameter; If the score representation value is greater than the second preset score representation value, it is determined that the generation parameter of the exclusive question set is abnormal, and the generation parameter of the exclusive question set for the single user is adjusted based on the wrong question dispersion.

[0030] In this embodiment, preferably, the first preset score representation value Z1 is selected within the interval [10, 12], and the second preset score representation value Z2 is selected within the interval [20, 25].

[0031] Specifically, the generation parameter of the exclusive question set for the single user is adjusted based on the wrong question dispersion, including: The time node at which the single user initially learns the knowledge points corresponding to each anchor point is obtained; Solving the time interval of each time node, solving the average value of each time interval, obtaining the wrong question discrete quantity; If the wrong question discrete quantity is less than or equal to the preset wrong question discrete quantity, the generation parameters of the exclusive question set for the single user are adjusted based on the data integrity of the current regular test paper; If the wrong question discrete quantity is greater than the preset wrong question discrete quantity, the number of selected random knowledge points is determined based on the wrong question discrete quantity.

[0032] Specifically, the specific way of determining the time node of the single user's initial learning of the knowledge points corresponding to each anchor point is not limited, which can be the time of obtaining the user's initial click on the corresponding course, or the time of the initial appearance of the corresponding knowledge point in the regular test paper, which will not be repeated here.

[0033] Specifically, the determination of the random knowledge points is to randomly select a number of knowledge points in the test question database, which will not be repeated here.

[0034] Specifically, the preset wrong question discrete quantity S0 is selected in the interval [30 days, 40 days].

[0035] Specifically, the number of added random knowledge points is determined based on the wrong question discrete quantity, wherein, The increase range of the number of selected random knowledge points is proportional to the wrong question discrete quantity.

[0036] In this embodiment, optionally, The wrong question discrete quantity is compared with the first preset discrete quantity comparison threshold and the second preset discrete quantity comparison threshold; If the wrong question discrete quantity is less than or equal to the first preset discrete quantity comparison threshold, the number of selected random knowledge points is determined as 0.2 times the preset selected number; If the wrong question discrete quantity is less than or equal to the second preset discrete quantity comparison threshold and greater than the first preset discrete quantity comparison threshold, the number of selected random knowledge points is determined as 0.3 times the preset selected number; If the wrong question discrete quantity is greater than the second preset discrete quantity comparison threshold, the number of selected random knowledge points is determined as 0.4 times the preset selected number; The first preset discrete quantity comparison threshold is 1.3S0, and the second preset discrete quantity comparison threshold is 2S0.

[0037] Specifically, the wrong question discrete quantity is determined, which represents the knowledge point forgetting situation of the user. When the wrong question discrete quantity is greater than the preset wrong question discrete quantity, it is determined that the reason for the abnormal decrease of the user's score is knowledge point forgetting. At this time, the random knowledge points are increased to form the exclusive question set for the user, which effectively improves the pertinence of the exclusive question set.

[0038] Specifically, the number of the conventional scores used to determine the score representation value is not limited, and it can be understood that the number should be no less than 5 to obtain the score change of a single user.

[0039] Specifically, the first preset score representation value is selected based on a large amount of teaching data statistics and analysis, and it is found that when the variance is in this range, the learning state of the user is relatively stable, and the current exclusive question set generation parameter can meet the learning needs of the user. The determination of the second preset score representation value is by analyzing and researching the score fluctuation of different user groups, and it is found that when the variance is in this range, there may be external factors such as test paper difficulty. It can be understood that those skilled in the art can adjust the selection of the value according to the actual situation.

[0040] Specifically, the process of determining whether the generation parameter of the exclusive question set for a single user is qualified in combination with the transverse comparison parameter includes: Calculate the average value of each conventional score of the single user monitored continuously to obtain an average score; Calculate the ratio of the conventional score of the current conventional test paper to the average score of each conventional score of the single user to obtain a score difference parameter for the current conventional test paper; Obtain the score difference parameter of each user for the current conventional test paper, and solve the average value to obtain a rating benchmark; Solve the ratio of the score difference parameter to the rating benchmark to obtain a transverse comparison parameter; If the transverse comparison parameter is less than or equal to a preset transverse comparison parameter, it is determined that the generation parameter of the exclusive question set is abnormal, and the generation parameter of the exclusive question set for a single user is adjusted based on the error question dispersion quantity; If the transverse comparison parameter is greater than the preset transverse comparison parameter, the first preset score representation value is adjusted to the corresponding value based on the transverse comparison parameter.

[0041] In this embodiment, preferably, the preset transverse comparison parameter B0 is selected in the interval [0.8, 0.9], Specifically, the selection of the preset transverse comparison parameter is based on the transverse comparison and analysis of a large number of user test scores, and it is found that when the ratio is in this range, the user's score decline can be more accurately judged whether it is related to the generation parameter of the exclusive question set. It can be understood that those skilled in the art can adjust the selection of the value according to the actual situation.

[0042] Specifically, the generation parameters of the exclusive question set for a single user are determined to be qualified based on the score representation value. When the score representation value is less than or equal to the first preset score representation value, the individual user's performance fluctuation is small and is in a relatively stable state. At this time, the generation parameters of the exclusive question set are determined to be qualified. When the score representation value is less than or equal to the second preset score representation value and greater than the first preset score representation value, the user's performance has slightly fluctuated. Further analysis is performed in combination with the horizontal comparison parameter. The generation parameters of the exclusive question set for a single user are determined to be qualified in combination with the horizontal comparison parameter. The horizontal comparison parameter represents the performance fluctuation of a single user compared with other users in the same period. When the horizontal comparison parameter is less than or equal to the preset horizontal comparison parameter, the individual user's performance has shown a significant decline relative to the same group of users. In this case, the generation parameters of the exclusive question set are adjusted. When the horizontal comparison parameter is greater than the preset horizontal comparison parameter, the performance of users in the same period has generally declined. It is determined that the score fluctuation is large due to the abnormally high difficulty of the test paper. At this time, the first preset score representation value is adjusted. The user's regular scores are continuously collected, and the variance of each regular score is calculated as the score representation value. Based on the comparison results between the score representation value and the preset score representation value, it is determined whether the generation parameters of the exclusive question set need to be adjusted. Fluctuations in user scores can be discovered in a timely manner, and the generation parameters of the exclusive question set can be reasonably adjusted according to different situations to ensure the effectiveness and pertinence of teaching. Combined with horizontal comparison parameters, it is determined whether the generation parameters of the exclusive question set for a single user are qualified. Taking into account the performance of users in the same group, it avoids misjudgments caused by external factors such as the difficulty of the test paper, and more accurately determines whether the generation parameters of the exclusive question set need to be adjusted. This improves teaching effectiveness and user learning efficiency, improves the efficiency of processing teaching information, and further improves the accuracy of user portraits.

[0043] Specifically, based on the data completeness of the current regular test paper, the generation parameters of the exclusive question set for a single user are adjusted, including: Obtain the data of each answer of the current regular test paper, identify the null values ​​in each data, calculate the ratio of the number of null values ​​to the number of preset test questions corresponding to the current regular test paper, and obtain the completeness of the data; the null values ​​are the unanswered questions in the test paper.

[0044] If the completeness is greater than the preset completeness, the number of test papers of the obtained historical regular test papers is adjusted to the corresponding value based on the score representation value; If the completeness is less than or equal to the first preset completeness, the generation parameters of the exclusive question set for the single user are adjusted based on the proportion of identified abnormal data.

[0045] Specifically, the first preset completeness W1 is selected within the interval [0.05, 0.1].

[0046] Specifically, the determination of the first preset completeness is based on statistics of the user's test paper answering situation. It is found that when the answering completeness is within this range, the proportion of insufficient knowledge point mastery of the user is large. It can be understood that the selection of the value can be adjusted by the person skilled in the art according to the actual situation.

[0047] Specifically, the specific way of obtaining the data of each answer of the current regular test paper, identifying the number of unanswered test questions and identifying abnormal data is not limited. The abnormal value is the data that cannot be clearly identified, which can include image preprocessing, binarization, character segmentation, character recognition, and character recognition using a deep learning model such as a convolutional neural network (CNN). The identification of null values and abnormal values includes determining that the character is not detected in the answering area after character segmentation as an unanswered test question. For characters that cannot be matched with the pre-defined character set, it is determined as abnormal data. This is prior art and will not be repeated.

[0048] Specifically, the preset number corresponding to the current regular test paper is the total number of areas that need to be answered in the test paper.

[0049] Specifically, the process of adjusting the generation parameters of the exclusive question set for a single user based on the proportion of identified abnormal data includes: Obtaining the data of each answer of the current regular test paper, identifying the number of abnormal data, calculating the ratio of the number of identified abnormal data to the preset number corresponding to the current regular test paper, and obtaining the proportion of identified abnormal data; If the proportion of identified abnormal data is less than or equal to the preset proportion of identified abnormal data, the generation parameters of the exclusive question set for a single user are adjusted based on the historical missing frequency of the single user; If the proportion of identified abnormal data is greater than the preset proportion of identified abnormal data, an alarm information for test paper content recognition is issued.

[0050] Specifically, the preset proportion of identified abnormal data is selected within the interval [0.2, 0.24].

[0051] Specifically, the preset proportion of identified abnormal data is analyzed by analyzing the accuracy and error rate of test paper content recognition. It is found that when the proportion of identified abnormal data is within this range, it can be better judged whether manual intervention is needed. It can be understood that the selection of the value can be adjusted by the person skilled in the art according to the actual situation.

[0052] Specifically, the generation parameters of the exclusive question set for a single user are adjusted based on the data integrity of the current conventional test paper. The integrity represents the completion of the test paper. When the integrity is greater than a first preset integrity, the user has completed most of the answers to the test paper. Fluctuations in the scores under this condition are determined to be weak knowledge points. The number of historical conventional test papers is adjusted at this time to more accurately determine the weak knowledge points of the user in the near future. When the data integrity is less than or equal to the first preset integrity, there may be situations where the answer cannot be accurately recognized due to messy handwriting, resulting in abnormal fluctuations in the score. At this time, the generation parameters of the exclusive question set are adjusted based on the proportion of abnormal data recognized. According to the user's answering situation and the test paper recognition situation, the generation parameters of the exclusive question set are flexibly adjusted, improving the adaptability and accuracy of teaching. The generation parameters of the exclusive question set for a single user are adjusted based on the proportion of abnormal data recognized. The abnormal data recognized represent the data that cannot be recognized or misrecognized. When the proportion of abnormal data recognized is greater than a preset proportion of abnormal data recognized, there is a large amount of data that cannot be recognized in the test paper, and an alarm information for test paper content recognition is issued. Problems in test paper content recognition are discovered in a timely manner to ensure the accuracy of the data, provide a reliable basis for subsequent teaching analysis and question set generation, improve the processing efficiency of teaching information, and further improve the accuracy of the user portrait.

[0053] Specifically, after the alarm information for test paper content recognition is issued, the handwritten information in the test paper of a single user is manually annotated. The inaccurate content is corrected. Then the annotated content is fed back to the training model used to recognize the content of the test paper. The model parameters are adjusted by increasing the training data to improve the recognition accuracy. The accuracy of test paper content recognition is continuously optimized to provide more accurate data support for subsequent teaching information processing.

[0054] Specifically, the process of adjusting the generation parameters of the exclusive question set for a single user based on the historical absence frequency of a single user includes: Obtain the ratio of the number of absences of a single user in a preset historical time period to the number of examinations of the single user in the historical time period to obtain the historical absence frequency of the single user; If the historical absence frequency is less than or equal to a preset historical absence frequency, delete the unanswered test questions in the historical conventional test paper used to determine the knowledge points corresponding to each wrong question; If the historical absence frequency is greater than the preset historical absence frequency, adjust the preset selected number to a corresponding value based on the integrity.

[0055] Specifically, the preset historical absence frequency is selected within the interval [0.2, 0.3].

[0056] Specifically, when the proportion of the abnormal data is less than or equal to the preset proportion of the abnormal data, the user's answer content can be clearly identified by the program. In this case, due to the absence of the user, the score is abnormally fluctuated. Based on the historical absence frequency of a single user, the generation parameters of the exclusive question set for the single user are adjusted. When the historical absence frequency is less than or equal to the preset historical absence frequency, it is determined that the single user is absent due to accidental and sudden phenomenon. At this time, the null value is removed when determining the knowledge points corresponding to the wrong questions, so as to improve the accuracy of the exclusive question set. When the historical absence frequency is greater than the preset historical absence frequency, the single user frequently absent in recent period. At this time, the preset selected number is adjusted to focus on the key problems of the single user. The influence of the absence of the user on the learning effect is considered, so that the generation of the exclusive question set is more in line with the actual situation of the user, the processing efficiency of the teaching information is improved, and the portrait accuracy for the user is further improved.

[0057] Specifically, the preset selected number is adjusted to a corresponding value based on the completeness, wherein, The reduction range of the preset percentage is inversely proportional to the completeness.

[0058] In this embodiment, optionally, The completeness is compared with a first preset completeness comparison threshold and a second preset completeness comparison threshold; If the completeness is less than or equal to the first preset completeness comparison threshold, the preset percentage is adjusted to 0.73 times of the initial preset percentage; If the completeness is less than or equal to the second preset completeness comparison threshold and greater than the first preset completeness comparison threshold, the preset percentage is adjusted to 0.83 times of the initial preset percentage; If the completeness is greater than the second preset completeness comparison threshold, the preset percentage is adjusted to 0.93 times of the initial preset percentage; The first preset completeness comparison threshold is 0.3W1, and the second preset completeness comparison threshold is 0.7W1.

[0059] Specifically, the first preset score representation value is adjusted to a corresponding value based on the lateral comparison parameter, wherein, The increase range of the first preset score representation value is proportional to the lateral comparison parameter.

[0060] In this embodiment, optionally, The lateral comparison parameter is compared with a first preset lateral comparison threshold and a second preset lateral comparison threshold; If the lateral comparison parameter is less than or equal to the first preset lateral comparison threshold, the first preset score representation value is adjusted to 1.11 times of the initial first preset score representation value; if the lateral comparison parameter is less than or equal to the second preset lateral comparison threshold and greater than the first preset lateral comparison threshold, the first preset score representation value is adjusted to 1.21 times of the initial first preset score representation value; if the lateral comparison parameter is greater than the second preset lateral comparison threshold, the first preset score representation value is adjusted to 1.31 times of the initial first preset score representation value; The first preset lateral comparison threshold is 1.5B0, and the second preset lateral comparison threshold is 1.8B0.

[0061] Specifically, when the adjustment for the first preset lateral comparison threshold is completed, the process of determining whether the generation parameter of the exclusive question set for the single user is qualified based on the adjusted first preset lateral comparison threshold, includes: if the score representation value is less than or equal to the adjusted first preset score representation value, it is determined that the generation parameter of the exclusive question set is qualified, and the current parameter is continuously used to complete the generation of the exclusive question set for the single user; if the score representation value is greater than the adjusted first preset score representation value, it is determined that the generation parameter of the exclusive question set is abnormal, and the generation parameter of the exclusive question set for the single user is adjusted based on the error question dispersion quantity.

[0062] Specifically, the number of historical regular test papers obtained is adjusted to a corresponding value based on the score representation value, wherein, The reduction range of the number of test papers is proportional to the score representation value.

[0063] In this embodiment, optionally, The score representation value is compared with the first preset score representation threshold and the second preset score representation threshold; if the score representation value is less than or equal to the first preset score representation threshold, the number of historical regular test papers obtained is adjusted to 0.92 times of the initial number of test papers; if the score representation value is less than or equal to the second preset score representation threshold and greater than the first preset score representation threshold, the number of historical regular test papers obtained is adjusted to 0.82 times of the initial number of test papers; if the score representation value is greater than the second preset score representation threshold, the number of historical regular test papers obtained is adjusted to 0.72 times of the initial number of test papers; The first preset score representation threshold is 2.4Z2, and the second preset score representation threshold is 3.7Z2.

[0064] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after the changes or replacements will all fall within the protection scope of the present application.

[0065] The above only describes the preferred embodiments of the present application and is not intended to limit the present application; the present application can have various changes and variations for those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A teaching information processing method based on AI, characterized in that: include: Determine the knowledge points corresponding to each characteristic character segment based on the characteristic character segments of a single user in the recognized conventional database; Arrange each knowledge point in descending order according to the frequency of occurrence, and select a preset number of knowledge points as anchor points; Selecting a number of questions corresponding to the anchor points from the test question database based on the anchor points to generate a unique question set for a single user; Determine a score representation value based on a regular score, wherein the regular score is a score of a regular test paper of a single user; Determining whether generation parameters of a data set for a single user are qualified based on the score representation value; When it is determined that the generation parameters of the dataset are abnormal, the generation parameters of the dataset for a single user are adjusted based on the discrete amount of wrong questions, including adding random knowledge points as anchor points; Determine whether the generation parameters of the dataset are qualified, and continue to use the current parameters to complete the generation of the dataset for a single user.

2. The AI-based teaching information processing method according to claim 1, characterized in that: The process of determining whether generation parameters of a data set for a single user are qualified based on the score representation value includes: When the score representation value is less than or equal to the second preset score representation value and greater than the first preset score representation value, determining whether the generation parameters of the data set for the single user are qualified in combination with the horizontal comparison parameters; Determine horizontal comparison parameters based on the regular scores of each user that are continuously monitored; When the horizontal comparison parameter is less than or equal to the preset horizontal comparison parameter, the generation parameters of the data set are determined to be abnormal, and the generation parameters of the data set for a single user are adjusted based on the discrete amount of wrong questions.

3. The AI-based teaching information processing method according to claim 2, characterized in that: When the horizontal comparison parameter is greater than a preset horizontal comparison parameter, the first preset score representation value is adjusted to a corresponding value based on the horizontal comparison parameter.

4. The AI-based teaching information processing method according to claim 2, characterized in that: When the score representation value is greater than the second preset score representation value, it is determined that the generation parameters of the data set are abnormal, and the generation parameters of the data set for the single user are adjusted based on the discrete amount of wrong questions.

5. The AI-based teaching information processing method according to claim 4, characterized in that: Adjusting generation parameters of a data set for a single user based on the discrete amount of wrong questions includes: Determine the discrete number of wrong questions based on the time nodes when a single user first learns the knowledge points corresponding to each anchor point; When the discrete number of wrong questions is less than or equal to the preset discrete number of wrong questions, the generation parameters of the data set for the single user are adjusted based on the data completeness of the current regular test paper; When the discrete number of wrong questions is greater than a preset discrete number of wrong questions, the number of random knowledge points to be selected is determined based on the discrete number of wrong questions.

6. The AI-based teaching information processing method according to claim 4, characterized in that: Adjust the generation parameters of the dataset for a single user based on the data completeness of the current regular test paper, including: Based on the data of each answer of the current regular test paper obtained, identifying the null values ​​in each data, and determining the completeness of the data based on the number of null values; When the completeness is greater than a preset completeness, the number of historical regular test papers in the regular database obtained for the single user is adjusted to a corresponding value based on the score representation value.

7. The AI-based teaching information processing method according to claim 6, characterized in that: When the completeness is less than or equal to a first preset completeness, adjusting generation parameters of a data set for a single user based on the identified abnormal data proportion includes: Based on the data of each answer of the current regular test paper obtained, the number of abnormal data is identified, and the proportion of abnormal data is determined based on the number of identified abnormal data; When the proportion of identified abnormal data is greater than the preset proportion of identified abnormal data, an alarm message for test paper content recognition is issued.

8. The AI-based teaching information processing method according to claim 7, characterized in that: When the proportion of identified abnormal data is less than or equal to the preset proportion of identified abnormal data, the generation parameters of the data set for the individual user are determined and adjusted based on the historical frequency of absence of the individual user, including: The historical absence frequency of the single user is determined based on the acquired number of absences of the single user within a preset historical period.

9. The AI-based teaching information processing method according to claim 8, characterized in that: When the historical absence frequency is less than or equal to the preset historical absence frequency, the empty values ​​in the history regular test paper used to determine the knowledge points corresponding to each characteristic character segment will be deleted.

10. The AI-based teaching information processing method according to claim 9, characterized in that: When the historical absence frequency is greater than the preset historical absence frequency, the preset selected number is adjusted to a corresponding value based on the completeness.

Images