Method, device and equipment for pushing on-line education training hall test questions

Abstract

The invention discloses a method, a device and equipment for pushing on-site test questions of online education and training, and relates to the technical field of artificial intelligence. The method comprises the steps of automatically extracting a current classroom new learning knowledge point set on the basis of a current classroom lecture text and a historical classroom lecture text, generating a plurality of corresponding classroom test questions for each new learning knowledge point in the set, and sending the test questions to the classroom. And according to the class-following monitoring video segment of which the acquisition time period is consistent with the display time period of the handout page to which the corresponding knowledge point belongs, obtaining the confidence coefficient of carefully attending the class on the corresponding knowledge point of the target student, and determining the corresponding question setting amount and / or question setting difficulty which are / is negatively related to the confidence coefficient, thereby obtaining the class-following monitoring video segment of which the acquisition time period is consistent with the display time period of the handout page to which the corresponding knowledge point belongs. And finally, according to the question setting amount and / or the question setting difficulty, at least one in-class test question is extracted and pushed to the target student, so that not only can the in-class test question be extracted quickly and efficiently, but also the question setting test paper can be ensured to have individual pertinence to the student, and the question setting efficiency and quality are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and particularly relates to a method, device and equipment for pushing in-class test questions in an online education training camp. Background Art

[0002] As is well known, the learning effect of a group is far greater than that of an individual. The online education training camp activities launched by training institutions are a new online hybrid training model of "online knowledge input + community service". With the online community as the carrier, gathering students with the same learning needs can not only efficiently and systematically complete learning tasks, but also create more opportunities for communication and sharing, greatly enhancing the learning experience.

[0003] Currently, after completing the training camp courses, in-class tests are generally required. The in-class test questions are mainly set by teachers or teaching teams, and then corresponding in-class test papers are formed. This process is time-consuming and laborious. In addition, the test papers are for all students and lack individual pertinence for students.

[0004] With the continuous development of artificial intelligence technology, how to use AI (Artificial Intelligence) technology to automatically generate personalized in-class test questions to improve the efficiency and quality of question setting is an urgent research topic for those skilled in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, device, computer equipment, computer-readable storage medium and computer program product for pushing in-class test questions in an online education training camp, so as to solve the problems of low efficiency of manual question setting and lack of individual pertinence for students in the existing in-class test question setting scheme.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] In a first aspect, a method for pushing in-class test questions in an online education training camp is provided, including:

[0008] Obtaining the current in-class lecture notes text and historical in-class lecture notes text of the online education training camp teaching course;

[0009] Structuring the current in-class lecture notes text to obtain a structured text of the current in-class lecture notes, and also structuring the historical in-class lecture notes text to obtain a structured text of the historical in-class lecture notes;

[0010] Performing knowledge point extraction processing on the structured text of the current in-class lecture notes to obtain a set of current in-class knowledge points, and also performing knowledge point extraction processing on the structured text of the historical in-class lecture notes to obtain a set of historical in-class knowledge points;

[0011] Take the remaining set after removing the intersection of the current in-class knowledge point set and the historical in-class knowledge point set from the current in-class knowledge point set as the current in-class newly learned knowledge point set;

[0012] For each newly learned knowledge point in the current in-class newly learned knowledge point set, generate a corresponding number of in-class test questions;

[0013] For each of the newly learned knowledge points, extract a segment of the in-class listening monitoring video from the current in-class listening monitoring video of the target student, where the acquisition period is consistent with the display period of the corresponding lecture page of the knowledge point, and import this segment of the in-class listening monitoring video into a pre-trained listening state recognition model based on the first artificial intelligence algorithm to output the first confidence level that the target student's listening state for the corresponding knowledge point is a serious listening state;

[0014] For each of the newly learned knowledge points, determine the corresponding question quantity and / or question difficulty that is negatively correlated with the confidence level according to the corresponding first confidence level, and extract at least one in-class test question from the corresponding multiple in-class test questions according to the question quantity and / or the question difficulty and push it to the target student.

[0015] Based on the above invention content, a new solution for automatically generating and pushing in-class test questions based on in-class lecture notes and in-class listening monitoring videos is provided. That is, first, based on the current in-class lecture note text and historical in-class lecture note text of the online education training camp teaching course, automatically extract the current in-class newly learned knowledge point set. Then, for each newly learned knowledge point in this set, generate a corresponding number of in-class test questions, and according to the segment of the in-class listening monitoring video where the acquisition period is consistent with the display period of the corresponding lecture page of the knowledge point, obtain the confidence level that the target student's listening state for the corresponding knowledge point is a serious listening state, and determine the corresponding question quantity and / or question difficulty that is negatively correlated with this confidence level. Finally, extract at least one in-class test question from the corresponding multiple in-class test questions according to the question quantity and / or the question difficulty and push it to the target student. In this way, not only can in-class test questions be extracted quickly and efficiently, but also the test papers generated can be targeted at individual students, improving the question generation efficiency and quality, and facilitating practical application and promotion.

[0016] In a possible design, perform knowledge point extraction processing on the current in-class lecture note structured text to obtain the current in-class knowledge point set, including:

[0017] Apply the extraction-based Q&A model LangChain to extract the first knowledge point set from the current structured class lecture text, and apply the first large language model ChatGPT based on the Transformer structure and pre-trained to generate the second knowledge point set according to the current structured class lecture text;

[0018] Match to obtain the knowledge point intersection of the first knowledge point set and the first knowledge point set;

[0019] Apply the second large language model RoBERTa pre-constructed and pre-trained based on natural language inference technology to determine whether there are semantically unsmooth parts in each knowledge point in the knowledge point intersection;

[0020] Use the knowledge point intersection with the knowledge points with semantically unsmooth parts removed as the current class knowledge point set.

[0021] In a possible design, for each newly learned knowledge point in the current class newly learned knowledge point set, generate corresponding multiple in-class test questions, including:

[0022] Apply semantic matching technology to select a predefined question type template suitable for each newly learned knowledge point in the current class newly learned knowledge point set. Among them, the predefined question type template includes a question stem slot and an answer option slot, and the question types of the predefined question type template cover multiple-choice questions, fill-in-the-blank questions, short-answer questions, true-false questions, and / or programming questions;

[0023] For each newly learned knowledge point, insert the corresponding knowledge point content into the question stem slot of the corresponding predefined question type template to obtain the corresponding question stem content. Then, apply a language generation model based on the GPT series language model to generate answer options that meet the requirements of grammar and semantics according to the question stem content, and insert the answer options into the answer option slot of the corresponding predefined question type template to obtain the corresponding answer content;

[0024] For each newly learned knowledge point, assemble the corresponding question stem content and answer content to obtain the corresponding in-class test question.

[0025] In a possible design, when the number of target students is multiple, for each newly learned knowledge point, extract at least one in-class test question from the corresponding multiple in-class test questions according to the corresponding number of questions to be set and push it to the target students, including:

[0026] For each of the newly learned knowledge points, traverse each of the target students in sequence: According to the corresponding number of questions K determined for the currently traversed student, randomly select K untaken classroom test questions from the corresponding multiple classroom test questions, and push the extraction result to the currently traversed student, where K represents a positive integer.

[0027] In a second aspect, there is provided a device for pushing classroom test questions in an online education training camp, including a lecture text acquisition unit, a structured processing unit, a knowledge point extraction unit, a newly learned knowledge point determination unit, a test question generation unit, a listening state classification unit, and a test question extraction unit that are sequentially communicatively connected;

[0028] The lecture text acquisition unit is used to acquire the current classroom lecture text and the historical classroom lecture text of the teaching course in the online education training camp;

[0029] The structured processing unit is used to perform structured processing on the current classroom lecture text to obtain a structured text of the current classroom lecture, and also perform structured processing on the historical classroom lecture text to obtain a structured text of the historical classroom lecture;

[0030] The knowledge point extraction unit is used to perform knowledge point extraction processing on the structured text of the current classroom lecture to obtain a set of current classroom knowledge points, and also perform knowledge point extraction processing on the structured text of the historical classroom lecture to obtain a set of historical classroom knowledge points;

[0031] The newly learned knowledge point determination unit is used to use the remaining set after removing the intersection of the current classroom knowledge point set and the historical classroom knowledge point set from the current classroom knowledge point set as the current classroom newly learned knowledge point set;

[0032] The test question generation unit is used to generate corresponding multiple classroom test questions for each of the newly learned knowledge points in the current classroom newly learned knowledge point set;

[0033] The listening state classification unit is used to, for each of the newly learned knowledge points, extract a segmented classroom listening monitoring video from the current classroom listening monitoring video of the target student, where the acquisition period of the segmented video is consistent with the display period of the lecture page to which the corresponding knowledge point belongs, and import the segmented classroom listening monitoring video into a listening state recognition model that is based on a first artificial intelligence algorithm and has been pre-trained, and output a first confidence level that the listening state of the target student on the corresponding knowledge point is a serious listening state;

[0034] The test question extraction unit is configured to, for each of the newly learned knowledge points, determine the corresponding question quantity and / or question difficulty that is negatively correlated with the confidence level according to the corresponding first confidence level, and extract at least one in-class test question from the corresponding multiple in-class test questions according to the question quantity and / or the question difficulty, and push the question to the target student.

[0035] In a third aspect, the present invention provides a computer device, including a memory, a processor, and a transceiver that are communicatively connected in sequence. Among them, the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the in-class test question pushing method of the online education training camp as described in the first aspect or any possible design in the first aspect.

[0036] In a fourth aspect, the present invention provides a computer-readable storage medium, on which instructions are stored. When the instructions are run on a computer, the in-class test question pushing method of the online education training camp as described in the first aspect or any possible design in the first aspect is executed.

[0037] In a fifth aspect, the present invention provides a computer program product, including a computer program or instructions. When the computer program or the instructions are executed by a computer, the in-class test question pushing method of the online education training camp as described in the first aspect or any possible design in the first aspect is implemented.

[0038] The beneficial effects of the above solution:

[0039] (1) The present invention creatively provides a new solution for automatically generating and pushing in-class test questions based on in-class lecture notes and in-class listening monitoring videos. That is, first, based on the current in-class lecture note text and historical in-class lecture note text of the online education training camp teaching course, the current set of newly learned knowledge points is automatically extracted. Then, for each newly learned knowledge point in the set, a corresponding multiple in-class test questions are generated, and according to the segmented in-class listening monitoring video where the acquisition period is consistent with the display period of the corresponding knowledge point's lecture page, the confidence level that the target student's listening state for the corresponding knowledge point is a serious listening state is obtained, and the corresponding question quantity and / or question difficulty that is negatively correlated with the confidence level is determined. Finally, at least one in-class test question is extracted from the corresponding multiple in-class test questions according to the question quantity and / or the question difficulty and pushed to the target student. In this way, not only can in-class test questions be quickly and efficiently extracted, but also the test paper can be ensured to be targeted at individual students, improving the question-setting efficiency and quality, and facilitating practical application and promotion. Description of the Drawings

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0041] Figure 1 It is a schematic flowchart of the method for pushing in-class test questions in the online education training camp provided by the embodiments of the present application.

[0042] Figure 2 It is a schematic structural diagram of the device for pushing in-class test questions in the online education training camp provided by the embodiments of the present application.

[0043] Figure 3 It is a schematic structural diagram of the computer device provided by the embodiments of the present application. Detailed implementation manners

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the present invention in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other embodiments can be obtained based on these embodiments. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention.

[0045] It should be understood that although terms such as first and second etc. may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another. For example, the first object can be called the second object, and similarly, the second object can be called the first object, without departing from the scope of the exemplary embodiments of the present invention.

[0046] It should be understood that for the term "and / or" that may appear in this article, it is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, B exists alone, or A and B exist simultaneously, etc.; and for example, A, B and / or C can represent any one of A, B and C or any combination of them; for the term " / and" that may appear in this article, it is a description of another association object relationship, indicating that two relationships may exist. For example, A / and B can represent: A exists alone or A and B exist simultaneously, etc.; in addition, for the character " / " that may appear in this article, generally it means that the associated objects before and after are in an "or" relationship.

[0047] Example:

[0048] As Figure 1 shown, the method for pushing current classroom test questions in the online education training camp provided in the first aspect of this embodiment can be, but is not limited to, executed by a computer device with certain computing resources, such as a platform server, a personal computer (Personal Computer, PC, referring to a multi-purpose computer suitable for personal use in terms of size, price, and performance; desktop computers, laptops, small laptops, tablet computers, and ultrabooks, etc. all belong to personal computers), a smart phone, a personal digital assistant (Personal Digital Assistant, PDA), or a wearable device and other electronic devices. As Figure 1 shown, the method for pushing current classroom test questions in the online education training camp can be, but is not limited to, including the following steps S1 to S7.

[0049] S1. Obtain the current classroom lecture text and the historical classroom lecture text of the online education training camp teaching course.

[0050] In step S1, the online education training camp teaching course is the teaching subject in the online education training camp teaching activity carried out for distributed students with the help of the Internet, such as a programming subject, etc. The current classroom lecture text is the teaching material explained in this online class, and the historical classroom lecture text is the teaching material explained in the previous and several previous online classes, and they can all be obtained regularly through the upload method of teaching staff.

[0051] S2. Perform structured processing on the current classroom lecture text to obtain a structured text of the current classroom lecture, and also perform structured processing on the historical classroom lecture text to obtain a structured text of the historical classroom lecture.

[0052] In step S2, the structured text is also a line-based text file, which is a very common file format (such as TXT format or CSV format, etc.), and usually each line corresponds to a record, and each line has the same columns, which is equivalent to a data table in a database. In order to be able to complete the generation and pushing of classroom test questions subsequently, it is first necessary to convert the current / historical classroom lecture text, which is human language, into a form that can be understood by a computer, so as to help the computer understand the meaning, structure, and grammar of the text, and extract useful information from it. Specifically, performing structured processing on the current classroom lecture text to obtain a structured text of the current classroom lecture includes, but is not limited to, the following steps S21 to S23.

[0053] S21. Use natural language processing technology to perform grammar parsing processing on the current classroom lecture text to obtain grammar information including part of speech and sentence structure.

[0054] In the step S21, the natural language processing technology can, but is not limited to, specifically adopt existing Tokenize technology (i.e., word segmentation technology) and / or Part-of-speech tagging technology (i.e., part-of-speech tagging technology), etc.

[0055] S22. Identify at least one phrase representing key knowledge points from the grammar information by using a rule / template matching method.

[0056] In the step S22, the at least one phrase can, but is not limited to, be exemplified as a noun phrase and / or a verb phrase, etc.

[0057] S23. Use a predefined knowledge system template / ontology to determine the key knowledge points represented by each phrase in the at least one phrase, and through structured representation, obtain the structured text of the current classroom lecture.

[0058] In the step S2, based on the foregoing steps S21 to S23, the technical details of obtaining the structured text of the historical classroom lecture by performing the structured processing on the historical classroom lecture text can also be deduced, which will not be elaborated here.

[0059] S3. Perform knowledge point extraction processing on the structured text of the current classroom lecture to obtain a set of current classroom knowledge points, and also perform knowledge point extraction processing on the structured text of the historical classroom lecture to obtain a set of historical classroom knowledge points.

[0060] In the step S3, the set of current classroom knowledge points includes all the knowledge points explained in the current online classroom. Specifically, in addition to the newly explained knowledge points, it may also include the knowledge points explained historically in the previous and several previous online classrooms (i.e., repeatedly explained knowledge points). The set of historical classroom knowledge points includes all the knowledge points explained in the previous and several previous online classrooms. To achieve the purpose of using AI (Artificial Intelligence) technology to complete knowledge point extraction and self-check whether there are self-consistency problems with the knowledge points, preferably, the knowledge point extraction processing on the structured text of the current classroom lecture to obtain the set of current classroom knowledge points includes, but is not limited to, the following steps S31 to S34.

[0061] S31. Apply the extraction-based question-answering model LangChain to extract a first set of knowledge points from the structured text of the current classroom lecture, and apply the first large language model ChatGPT based on the Transformer structure and pre-trained to generate a second set of knowledge points according to the structured text of the current classroom lecture.

[0062] In the step S31, the extraction-based question-answering model LangChain is a framework for developing applications based on language models, and it can implement the following applications: data awareness, that is, connecting the language model to other data sources; autonomy, that is, allowing the language model to interact with its environment, so the first knowledge point set can be routinely extracted from the current structured text of the lecture notes. The first large language model ChatGPT (full name: Chat Generative Pre-trained Transformer, a chatbot program developed by OpenAI, released on November 30, 2022, which is a natural language processing tool driven by artificial intelligence technology. It can generate answers based on the patterns and statistical rules seen in the pre-training stage, and can also interact according to the context of the chat, truly chatting and communicating like a human, and even can complete tasks such as writing emails, video scripts, copywriting, translation, code, or writing papers) can but is not limited to specifically using the GPT-4 model.

[0063] S32. Match to obtain the knowledge point intersection of the first knowledge point set and the first knowledge point set.

[0064] S33. Apply the second large language model RoBERTa that is pre-constructed based on natural language inference technology and has completed pre-training to determine whether there are semantically unsmooth parts in each knowledge point in the knowledge point intersection.

[0065] In the step S33, the second large language model RoBERTa is an optimized version of the BERT (Bidirectional Encoder Representations from Transformers, bidirectional encoder representations from Transformer, a language representation model) model. Its main advantage is that it uses an online dynamic masking technique during the pre-training process, rather than performing the masking operation in an offline mode like BERT. In this way, no additional storage space is required, and the mask replacement is performed in real time during the model training, improving the efficiency and accuracy. Thus, based on natural language inference technology and a certain amount of positive and negative knowledge point samples, the second large language model RoBERTa that can effectively identify whether there are semantically unsmooth parts in the knowledge points can be routinely trained.

[0066] S34. Use the knowledge point intersection from which the knowledge points with semantically unsmooth parts have been removed as the current in-class knowledge point set.

[0067] In the step S3, based on the foregoing steps S31 - S34, the technical details of obtaining the historical in - class knowledge point set by performing knowledge point extraction on the structured text of the historical in - class lecture notes can also be deduced, which will not be elaborated here.

[0068] S4. Use the remaining set after removing the intersection of the current in - class knowledge point set and the historical in - class knowledge point set from the current in - class knowledge point set as the current in - class newly learned knowledge point set.

[0069] In the step S4, for example, if the current in - class knowledge point set contains knowledge points A, B, C, D, E, and F, and the historical in - class knowledge point set contains knowledge points A, B, D, and E, then the remaining set contains knowledge points C and F, which are all newly learned knowledge points in the current class.

[0070] S5. For each newly learned knowledge point in the current in - class newly learned knowledge point set, generate corresponding multiple in - class test questions.

[0071] In the step S5, in order to achieve the purpose of using AI technology to generate multiple in - class test questions of different question types, preferably, for each newly learned knowledge point in the current in - class newly learned knowledge point set, generating corresponding multiple in - class test questions includes but is not limited to the following steps S51 - S53.

[0072] S51. Apply semantic matching technology to select predefined question type templates suitable for each newly learned knowledge point in the current in - class newly learned knowledge point set. Among them, the predefined question type templates include but are not limited to question stem slots, answer option slots, etc. The question types covered by the predefined question type templates include but are not limited to multiple - choice questions, fill - in - the - blank questions, essay questions, true - or - false questions, and / or programming questions, etc.

[0073] S52. For each newly learned knowledge point, insert the corresponding knowledge point content into the question stem slot of the corresponding predefined question type template to obtain the corresponding question stem content. Then, apply a language generation model based on the GPT series language model to generate answer options that meet the requirements of grammar and semantics according to the question stem content, and insert the answer options into the answer option slot of the corresponding predefined question type template to obtain the corresponding answer content.

[0074] In the step S52, the GPT series language model can but is not limited to specifically adopt the GPT - 4 model.

[0075] S53. For each newly learned knowledge point, assemble the corresponding question stem content and answer content to obtain the corresponding in - class test question.

[0076] After step S53, in order to verify the usability of the in-class test questions, that is, to verify whether the in-class test questions contain conceptual errors (for example, the question itself is wrong), factual errors (for example, the correct answer options are already reflected in the question stem), logical errors (for example, there are no correct answer options) and / or grammatical errors (for example, the question description is incomplete or unclear), etc., it is further preferred that after obtaining the in-class test questions, the method further includes but is not limited to the following steps S541 to S543.

[0077] S541. Apply a test question error recognition model that is based on a second artificial intelligence algorithm and has completed pre-training to provide a second confidence level for whether the in-class test questions contain errors, wherein the test question error recognition model is used to identify whether the test questions contain errors, and the errors include but are not limited to conceptual errors, factual errors, logical errors and / or grammatical errors, etc.

[0078] In the step S541, the artificial intelligence algorithm is a core artificial intelligence algorithm that specifically studies how computers simulate or implement human learning behavior to acquire new knowledge or skills, reorganize existing knowledge structures and continuously improve their own performance. It is the fundamental way to make computers intelligent; specifically, the second artificial intelligence algorithm can be implemented by, but is not limited to, machine learning algorithms such as support vector machines, K-nearest neighbor method, stochastic gradient descent method, multi-layer perceptron, decision tree, back propagation neural network or radial basis function network, and can be pre-trained to obtain the test question error recognition model through conventional calibration verification modeling method (the specific process includes the calibration process and verification process of the model, that is, first comparing the model simulation results with the measured data, and then adjusting the model parameters according to the comparison results to make the simulation results consistent with the actual process).

[0079] S542. Determine whether the second confidence level exceeds a preset confidence level threshold.

[0080] In step S542 , the confidence threshold may be 61.8% for example.

[0081] S543. If yes, remove the in-class test question.

[0082] S6. For each newly learned knowledge point, extract a segment of the target student's current in-class monitoring video, the acquisition period of which is consistent with the display period of the handout page corresponding to the knowledge point, and import the segment into a pre-trained listening state recognition model based on a first artificial intelligence algorithm, and output a first confidence level that the target student's listening state at the corresponding knowledge point is a serious listening state.

[0083] In the step S6, the target student is the object to be evaluated during the current online class, and the current in-class monitoring video is the monitoring video regularly recorded for the listening performance of the target student in the current online class (which can be regularly recorded by the camera on the student side and uploaded to the local). For example, if the display period of the lecture page of a newly learned knowledge point in the current online class is from 10:00 to 10:05, then for the newly learned knowledge point, it is necessary to extract a segment of the in-class monitoring video with a time period between 10:00 and 10:05 from the current in-class monitoring video of the target student. Since the segment of the in-class monitoring video will present the micro-expression features of the target student during the display period of the lecture page: behavioral features such as concentration or distraction, etc., the listening state recognition model can be pre-trained through a conventional calibration and verification modeling method based on a certain amount of samples of the in-class monitoring video segments and sample labels (such as the state of listening attentively or not listening attentively). In addition, specifically, the first artificial intelligence algorithm can but is not limited to adopting machine learning algorithms based on support vector machines, decision trees, or random forests.

[0084] S7. For each newly learned knowledge point, determine the corresponding question quantity and / or question difficulty that is negatively correlated with the confidence level according to the corresponding first confidence level, and extract at least one in-class test question from the corresponding multiple in-class test questions according to the question quantity and / or the question difficulty, and push it to the target student.

[0085] In the step S7, the purpose of establishing the negative correlation between the first confidence level and the number of questions and / or the difficulty of questions is as follows: if the target student listens more attentively to a newly learned knowledge point (i.e., the first confidence level is higher), it can be reflected that the target student has a higher mastery level of this newly learned knowledge point. Therefore, fewer questions or easier questions can be pushed for the in-class routine test. If the target student listens less attentively to a newly learned knowledge point (i.e., the first confidence level is lower), it can be reflected that the target student has a lower mastery level of this newly learned knowledge point. Therefore, more questions or more difficult questions can be pushed for the in-class focused test, so as to improve the target student's mastery level of this newly learned knowledge point. Furthermore, through this targeted question-setting method, the mastery levels of all students for the newly learned knowledge points can be balanced, which is conducive to all of them keeping up with the subsequent online education training camp teaching courses and improving the activity effect of the online education training camp. In addition, considering that the students in the training camp are not concentrated together for teaching and training, there is a problem of difficult test supervision. Therefore, to avoid the phenomenon that multiple students copy answers from each other by means of Internet communication technology (such as through WeChat), preferably, when there are multiple target students, for each newly learned knowledge point, at least one in-class test question is extracted from the corresponding multiple in-class test questions according to the corresponding number of questions and pushed to the target students, including but not limited to: for each newly learned knowledge point, each of the target students is traversed in turn: according to the corresponding number of questions K determined for the currently traversed student, K unextracted in-class test questions are randomly selected from the corresponding multiple in-class test questions, and the extraction result is pushed to the currently traversed student, where K represents a positive integer. In this way, it can be ensured that the in-class test questions for each student on the same newly learned knowledge point are different, and thus it can effectively prevent them from copying answers from each other.

[0086] Based on the above-described method for pushing in-class test questions in the online education training camp in steps S1 to S7, a new solution for automatically generating and pushing in-class test questions based on the in-class lecture notes and in-class listening monitoring videos is provided. That is, first, based on the current in-class lecture note text and the historical in-class lecture note text of the online education training camp teaching course, the current set of newly learned knowledge points in the in-class is automatically extracted. Then, for each newly learned knowledge point in this set, a corresponding multiple in-class test questions are generated, and according to the segmented in-class listening monitoring videos where the acquisition period is consistent with the display period of the corresponding knowledge point's lecture page, the confidence level that the target student's listening state for the corresponding knowledge point is a serious listening state is obtained, and the corresponding number of questions and / or the difficulty of questions negatively correlated with this confidence level are determined. Finally, at least one in-class test question is extracted from the corresponding multiple in-class test questions according to this number of questions and / or this difficulty of questions and pushed to the target student. In this way, not only can the in-class test questions be quickly and efficiently extracted, but also the test papers can be ensured to be targeted at individual students, improving the question-setting efficiency and quality, and facilitating practical application and promotion.

[0087] As Figure 2 shown, in the second aspect of this embodiment, a virtual device for implementing the online education training camp in-class test question pushing method described in the first aspect is provided, including a lecture text acquisition unit, a structured processing unit, a knowledge point extraction unit, a newly learned knowledge point determination unit, a test question generation unit, a listening state classification unit, and a test question extraction unit that are sequentially communicatively connected;

[0088] The lecture text acquisition unit is used to acquire the current in-class lecture text and the historical in-class lecture text of the online education training camp teaching course;

[0089] The structured processing unit is used to perform structured processing on the current in-class lecture text to obtain the current in-class lecture structured text, and also perform structured processing on the historical in-class lecture text to obtain the historical in-class lecture structured text;

[0090] The knowledge point extraction unit is used to perform knowledge point extraction processing on the current in-class lecture structured text to obtain the current in-class knowledge point set, and also perform knowledge point extraction processing on the historical in-class lecture structured text to obtain the historical in-class knowledge point set;

[0091] The newly learned knowledge point determination unit is used to use the remaining set after removing the intersection of the current in-class knowledge point set and the historical in-class knowledge point set in the current in-class knowledge point set as the current in-class newly learned knowledge point set;

[0092] The test question generation unit is used to generate a corresponding plurality of in-class test questions for each newly learned knowledge point in the current in-class newly learned knowledge point set;

[0093] The listening state classification unit is used to, for each newly learned knowledge point, intercept a segment of the in-class listening monitoring video whose acquisition period is consistent with the display period of the corresponding lecture page from the current in-class listening monitoring video of the target student, and import this segment of the in-class listening monitoring video into a listening state recognition model that is based on the first artificial intelligence algorithm and has been pre-trained, and output the first confidence level that the listening state of the target student on the corresponding knowledge point is a serious listening state;

[0094] The test question extraction unit is used to, for each newly learned knowledge point, determine the corresponding question quantity and / or question difficulty that is negatively correlated with the confidence level according to the corresponding first confidence level, and extract at least one in-class test question from the corresponding plurality of in-class test questions according to the question quantity and / or the question difficulty and push it to the target student.

[0095] For the working process, working details, and technical effects of the aforementioned device provided in the second aspect of this embodiment, reference may be made to the online education training camp in-class test question pushing method described in the first aspect, which will not be elaborated here.

[0096] As Figure 3 shown, in the third aspect of this embodiment, a computer device for executing the online education training camp in-class test question pushing method described in the first aspect is provided, including a memory, a processor, and a transceiver that are sequentially communicatively connected. Among them, the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer programs and execute the online education training camp in-class test question pushing method described in the first aspect. Specifically, for example, the memory may include, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a flash memory, a first input first output (FIFO), and / or a first input last output (FILO), etc.; the processor may include, but is not limited to, a microprocessor of the STM32F105 series. In addition, the computer device may further include, but is not limited to, a power module, a display screen, and other necessary components.

[0097] For the working process, working details, and technical effects of the aforementioned computer device provided in the third aspect of this embodiment, reference may be made to the online education training camp in-class test question pushing method described in the first aspect, which will not be elaborated here.

[0098] In the fourth aspect of this embodiment, a computer-readable storage medium storing instructions including the online education training camp in-class test question pushing method described in the first aspect is provided, that is, instructions are stored on the computer-readable storage medium. When the instructions are run on a computer, the online education training camp in-class test question pushing method described in the first aspect is executed. Among them, the computer-readable storage medium refers to a carrier for storing data, which may include, but is not limited to, a floppy disk, an optical disc, a hard disk, a flash memory, a USB flash drive, and / or a memory stick, etc. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0099] For the working process, working details, and technical effects of the aforementioned computer-readable storage medium provided in the fourth aspect of this embodiment, reference may be made to the online education training camp in-class test question pushing method described in the first aspect, which will not be elaborated here.

[0100] The fifth aspect of this embodiment provides a computer program product, including a computer program or instructions, and the computer program or the instructions, when executed by a computer, implement the online education training camp in-class test question pushing method as described in the first aspect. Among them, the computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices.

[0101] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for pushing in-class test questions in an online education training camp, characterized in that, Including: Obtaining the current in-class lecture notes text and historical in-class lecture notes text of the online education training camp teaching courses; Performing structured processing on the current in-class lecture notes text to obtain the current in-class lecture notes structured text, and also performing structured processing on the historical in-class lecture notes text to obtain the historical in-class lecture notes structured text; Performing knowledge point extraction processing on the current in-class lecture notes structured text to obtain the current in-class knowledge point set, and also performing knowledge point extraction processing on the historical in-class lecture notes structured text to obtain the historical in-class knowledge point set; Taking the remaining set after removing the intersection of the current in-class knowledge point set and the historical in-class knowledge point set from the current in-class knowledge point set as the current in-class newly learned knowledge point set; Generating corresponding multiple in-class test questions for each newly learned knowledge point in the current in-class newly learned knowledge point set; For each newly learned knowledge point, intercepting a segment of the in-class listening monitoring video whose acquisition period is consistent with the display period of the corresponding knowledge point's lecture page from the current in-class listening monitoring video of the target student, and importing this segment of the in-class listening monitoring video into a pre-trained listening state recognition model based on the first artificial intelligence algorithm to output the first confidence level that the target student's listening state for the corresponding knowledge point is a serious listening state; For each newly learned knowledge point, determining the corresponding question quantity and / or question difficulty that is negatively correlated with the confidence level according to the corresponding first confidence level, and extracting at least one in-class test question from the corresponding multiple in-class test questions according to the question quantity and / or the question difficulty and pushing it to the target student.

2. The online education training camp in-class test question pushing method according to claim 1, wherein Performing structured processing on the current in-class lecture notes text to obtain the current in-class lecture notes structured text, including: Performing grammar parsing processing on the current in-class lecture notes text using natural language processing technology to obtain grammar information including part of speech and sentence structure; Using the rule / template matching method to identify at least one phrase representing key knowledge points from the grammar information; Using a predefined knowledge system template / ontology to determine the key knowledge points represented by each phrase in the at least one phrase, and obtaining the current in-class lecture notes structured text through structured representation.

3. The online education training camp in-class test question pushing method according to claim 1, wherein Performing knowledge point extraction processing on the current in-class lecture notes structured text to obtain the current in-class knowledge point set, including: Applying the extraction-based question-answering model LangChain to extract a first knowledge point set from the current in-class lecture notes structured text, and applying the first large language model ChatGPT based on the Transformer structure and pre-trained to generate a second knowledge point set according to the current in-class lecture notes structured text; Matching to obtain the knowledge point intersection of the first knowledge point set and the first knowledge point set; Applying the second large language model RoBERTa pre-constructed based on natural language inference technology and pre-trained to judge whether there are semantically unsmooth parts in each knowledge point in the knowledge point intersection; Use the intersection of the knowledge points after removing the semantically incoherent parts of the knowledge points as the current in-class knowledge point set.

4. The online education training camp in-class test question pushing method according to claim 1, characterized in that, For each newly learned knowledge point in the current in-class newly learned knowledge point set, generate corresponding multiple in-class test questions, including: Use semantic matching technology to select predefined question type templates suitable for each newly learned knowledge point in the current in-class newly learned knowledge point set. Among them, the predefined question type templates include question stem slots and answer option slots, and the question types of the predefined question type templates cover multiple-choice questions, fill-in-the-blank questions, short-answer questions, true-or-false questions, and / or programming questions; For each newly learned knowledge point, insert the corresponding knowledge point content into the question stem slot of the corresponding predefined question type template to obtain the corresponding question stem content. Then, use a language generation model based on the GPT series language model to generate answer options that meet the requirements of grammar and semantics according to the question stem content, and insert the answer options into the answer option slot of the corresponding predefined question type template to obtain the corresponding answer content; For each newly learned knowledge point, assemble the corresponding question stem content and answer content to obtain the corresponding in-class test question.

5. The method for pushing in-class test questions of an online education training camp according to claim 1, wherein When the number of target students is multiple, for each newly learned knowledge point, extract at least one in-class test question from the corresponding multiple in-class test questions according to the corresponding number of questions to be set and push it to the target students, including: For each newly learned knowledge point, traverse each of the target students in turn: According to the corresponding number of questions to be set K determined for the currently traversed student, randomly extract K unextracted in-class test questions from the corresponding multiple in-class test questions, and push the extraction result to the currently traversed student, where K represents a positive integer.

6. The online education training camp in-class test question push method according to claim 1, characterized in that The first artificial intelligence algorithm uses a machine learning algorithm based on support vector machine, decision tree or random forest.

7. An online education training camp in-class test question push device, characterized in that, It includes a lecture text acquisition unit, a structured processing unit, a knowledge point extraction unit, a newly learned knowledge point determination unit, a test question generation unit, a listening status classification unit, and a test question extraction unit that are sequentially communicatively connected; The lecture text acquisition unit is used to acquire the current in-class lecture text and historical in-class lecture text of the online education training camp teaching course; The structured processing unit is used to perform structured processing on the current in-class lecture text to obtain the current in-class lecture structured text, and also perform structured processing on the historical in-class lecture text to obtain the historical in-class lecture structured text; The knowledge point extraction unit is used to perform knowledge point extraction processing on the current in-class lecture structured text to obtain the current in-class knowledge point set, and also perform knowledge point extraction processing on the historical in-class lecture structured text to obtain the historical in-class knowledge point set; The newly learned knowledge point determination unit is used to use the remaining set after removing the intersection of the current in-class knowledge point set and the historical in-class knowledge point set in the current in-class knowledge point set as the current in-class newly learned knowledge point set; The test question generation unit is used to generate corresponding multiple in-class test questions for each newly learned knowledge point in the current in-class newly learned knowledge point set; The class-attending status classification unit is configured to, for each of the newly learned knowledge points, extract a segment of the in-class monitoring video of the target student, whose acquisition period is consistent with the display period of the corresponding lecture page of the knowledge point, from the current in-class monitoring video of the target student, and import the segment of the in-class monitoring video into a pre-trained in-class status recognition model based on a first artificial intelligence algorithm, and output a first confidence level that the class-attending status of the target student for the corresponding knowledge point is a serious class-attending status; The test question extraction unit is configured to, for each of the newly learned knowledge points, determine a corresponding question quantity and / or question difficulty that is negatively correlated with the confidence level according to the corresponding first confidence level, and extract at least one in-class test question from a plurality of corresponding in-class test questions according to the question quantity and / or the question difficulty, and push the question to the target student.

8. A computer device, characterized in that, It includes a memory, a processor, and a transceiver that are communicatively connected in sequence. Among them, the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the in-class test question pushing method for the online education training camp according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that , An instruction is stored on the computer-readable storage medium. When the instruction runs on a computer, it executes the in-class test question pushing method for the online education training camp according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program or instructions, characterized in that, The computer program or the instruction, when executed by a computer, implements the in-class test question pushing method for the online education training camp according to any one of claims 1 to 6.

Images