Classroom question and answer analysis method and device, intelligent equipment and storage medium

Abstract

The invention is suitable for the technical field of artificial intelligence, and provides a classroom question and answer analysis method and device, intelligent equipment and a storage medium, and the classroom question and answer analysis method comprises the steps: obtaining a classroom record segment in real time based on a classroom question and answer analysis task; converting each class real-record segment acquired in real time into real-record text segments, and inputting the real-record text segments into the question-answer analysis large model; the question and answer analysis large model is obtained by performing fine tuning training on a large language model by using a training set meeting a custom specification; and carrying out classroom question-answer analysis processing on each recorded text segment by utilizing the question-answer analysis large model until all recorded text segments corresponding to the classroom question-answer analysis task are processed, and generating a large model question-answer analysis result. The method does not need to analyze the complete classroom record after the end of the classroom, can effectively shorten the time of classroom question and answer analysis, improves the timeliness and accuracy of the classroom question and answer analysis, and achieves the precise question and answer extraction granularity.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a classroom question and answer analysis method, apparatus, intelligent device and storage medium. Background Art

[0002] In the context of the rapid development of educational informatization, the quality of classroom interaction has become one of the key indicators for evaluating teaching effectiveness. As an important form of interaction between teachers and students, classroom Q&A plays an irreplaceable role in promoting knowledge transfer, stimulating students' thinking, and evaluating teaching effectiveness. Effective classroom Q&A can not only provide instant feedback on students' learning status and help teachers adjust their teaching strategies, but also enhance students' sense of participation and learning motivation. Therefore, in-depth analysis of classroom Q&A is of great significance for improving teaching quality and optimizing course design.

[0003] However, most traditional classroom question-and-answer analysis methods rely on manual video viewing or teacher self-reporting, which is not only time-consuming and laborious, but also easily affected by subjective factors, making it difficult to achieve efficient and accurate analysis. In addition, although the existing automated question-and-answer extraction solutions based on simple machine models can reduce the manual burden to a certain extent, due to the complexity of classroom scenes and question-and-answer contexts, automated question-and-answer extraction solutions based on simple machine models often have difficulty accurately identifying and distinguishing questions and answers in the classroom, which can easily lead to misjudgments and missed judgments.

[0004] In view of this, how to improve the timeliness and accuracy of classroom question and answer analysis and improve the granularity of question and answer extraction are technical problems that need to be solved at present. Summary of the invention

[0005] The embodiments of the present application provide a classroom question and answer analysis method, apparatus, intelligent device and storage medium, which can improve the timeliness and accuracy of classroom question and answer analysis and enhance the granularity of question and answer extraction.

[0006] In a first aspect, an embodiment of the present application provides a classroom question and answer analysis method, comprising:

[0007] Based on the classroom question and answer analysis task, the classroom transcript segmentation is obtained in real time;

[0008] Each of the classroom transcript segments obtained in real time is converted into transcript text segments and then input into a large question-answer analysis model, wherein the large question-answer analysis model is obtained by fine-tuning a large language model using a training set that meets custom specifications;

[0009] The question-and-answer analysis big model is used to perform classroom question-and-answer analysis on each of the transcript text segments until all transcript text segments corresponding to the classroom question-and-answer analysis task are processed, thereby generating a big model question-and-answer analysis result.

[0010] In a possible implementation of the first aspect, the step of converting each of the classroom transcript segments acquired in real time into transcript text segments and then inputting them into the question-answer analysis model includes:

[0011] The recorded text is segmented according to a preset sequence length to obtain a plurality of sequences;

[0012] The several sequences are input into the question-answer analysis model in sequence.

[0013] In a possible implementation of the first aspect, the step of sequentially inputting the plurality of sequences into the question-answer analysis large model includes:

[0014] If the sequence length of the end sequence in the transcript text segment is less than the preset sequence length, determining whether the transcript text segment is the end segment;

[0015] If the transcript text segment is the end segment, the end sequence of the end segment is input into the question-answer analysis model;

[0016] If the recorded text segment is not the end segment, suspending the end sequence of the recorded text segment and monitoring the next recorded text segment;

[0017] The end sequence of the recorded text segment is merged with the next recorded text segment detected, and divided according to the preset sequence length to obtain a plurality of new sequences;

[0018] The new sequences are input into the question-answer analysis model in sequence.

[0019] In a possible implementation of the first aspect, the step of using the question-and-answer analysis big model to perform classroom question-and-answer analysis on each of the transcript text segments until all transcript text segments corresponding to the classroom question-and-answer analysis task are processed, and then generating a big model question-and-answer analysis result includes:

[0020] Extracting the question and answer elements in the transcript text segment, and determining whether the transcript text segment is the end segment;

[0021] If the transcript text segment is not the end segment, the question and answer elements extracted from the transcript text segment are used as the intermediate analysis result;

[0022] If the transcript text segment is the end segment, the question and answer elements extracted from the transcript text segment are merged with all the intermediate analysis results to generate a large model question and answer analysis result corresponding to the classroom question and answer analysis task.

[0023] In a possible implementation of the first aspect, after the step of generating a large model question-answering analysis result, the method further includes:

[0024] Obtaining the transcript corresponding to the classroom question and answer analysis task;

[0025] Using a rule-based classification algorithm to extract questions from the transcript text;

[0026] The questions in the large model question and answer analysis results are optimized according to the questions.

[0027] In a possible implementation of the first aspect, the step of optimizing the questions in the large model question and answer analysis result according to the questions includes:

[0028] Calculating the text similarity between the first question and the second question, wherein the first question includes the question in the large model question-answering analysis result, and the second question includes the question extracted from the transcript text by a rule-based classification algorithm;

[0029] Based on the text similarity, the questions in the large model question and answer analysis results are deduplicated and supplemented.

[0030] In a possible implementation of the first aspect, the step of fine-tuning the large language model using a training set that meets the custom specification to obtain the large question and answer analysis model includes:

[0031] Build a training set that meets custom specifications;

[0032] Constructing a fine-tuning task, wherein the fine-tuning task is used to instruct the large language model to learn to perform classroom question-and-answer analysis on the transcript text corresponding to the classroom transcript, and generate structured question-and-answer analysis elements;

[0033] Fine-tune the large language model based on the fine-tuning task and the training set until a large question-answering analysis model that completes the fine-tuning task is obtained.

[0034] In a second aspect, an embodiment of the present application provides a classroom question and answer analysis device, comprising:

[0035] A real-time segment acquisition unit, used to acquire real-time segmentation of classroom records based on classroom question-answer analysis tasks;

[0036] A segmented text input unit is used to convert each of the classroom transcript segments obtained in real time into transcript text segments and then input them into the question-answer analysis model, wherein the question-answer analysis model is obtained by fine-tuning the large language model using a training set that meets the custom specification;

[0037] The question and answer analysis processing unit is used to use the question and answer analysis big model to perform classroom question and answer analysis on each of the transcript text segments until all the transcript text segments corresponding to the classroom question and answer analysis task are processed, and then generate a big model question and answer analysis result.

[0038] In a third aspect, an embodiment of the present application provides an intelligent device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the classroom question and answer analysis method as described in the first aspect above is implemented.

[0039] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the classroom question and answer analysis method as described in the first aspect above is implemented.

[0040] In a fifth aspect, an embodiment of the present application provides a computer program product, which, when executed on a smart device, enables the smart device to execute the classroom question and answer analysis method as described in the first aspect above.

[0041] In an embodiment of the present application, the smart device obtains classroom transcript segments in real time based on a classroom question and answer analysis task, converts each of the classroom transcript segments obtained in real time into transcript text segments, and then inputs them into a question and answer analysis big model, and uses the question and answer analysis big model to perform classroom question and answer analysis on each of the transcript text segments until all transcript text segments corresponding to the classroom question and answer analysis task are processed, and then the big model question and answer analysis results are automatically generated. There is no need to wait for the end of the class to analyze the complete classroom transcript, and there is no need to rely on manual record analysis. This can effectively shorten the time for classroom question and answer analysis and improve the timeliness and accuracy of classroom question and answer analysis. At the same time, the question and answer analysis big model obtained by fine-tuning the big language model with a training set that meets custom specifications in the present application solution generates the big model question and answer analysis results, which is conducive to the granularity of accurate question and answer extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 is a flow chart for implementing the classroom question and answer analysis method provided in the embodiment of the present application;

[0044] Figure 2It is a specific implementation flow chart of sequentially inputting the sequences into the question-and-answer analysis model in the classroom question-and-answer analysis method provided in the embodiment of the present application;

[0045] Figure 3 It is a specific implementation flow chart of obtaining a large model of question and answer analysis in the classroom question and answer analysis method provided in the embodiment of the present application;

[0046] Figure 3.1 It is a schematic diagram of structured sample labels in the classroom question-and-answer analysis method provided in an embodiment of the present application;

[0047] Figure 3.2 is a schematic diagram of serialized sample labels in the classroom question-and-answer analysis method provided in an embodiment of the present application;

[0048] Figure 4 This is a specific implementation flow chart of step S103 in the classroom question and answer analysis method provided in the embodiment of the present application;

[0049] Figure 5 It is a specific implementation flow chart of optimizing the question and answer analysis results in the classroom question and answer analysis method provided in the embodiment of the present application;

[0050] Figure 6 is a structural block diagram of a classroom question-and-answer analysis device provided in an embodiment of the present application;

[0051] Figure 7 It is a schematic diagram of a smart device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0053] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0054] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0055] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0056] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0057] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0058] The classroom question-and-answer analysis method provided in the embodiment of the present application is applicable to smart devices of various types of data that need to perform question-and-answer analysis, and the smart devices may specifically include mobile phones, tablet computers, wearable devices, laptop computers, ultra-mobile personal computers (UMPCs), desktop computers, smart large screens, and servers, etc. The embodiment of the present application does not impose any restrictions on the specific types of smart devices.

[0059] Figure 1 The implementation process of the classroom question and answer analysis method provided in the embodiment of the present application is shown. In the embodiment of the present application, the method flow includes steps S101 to S103. The specific implementation principle of each step is as follows:

[0060] Step S101: Based on the classroom question and answer analysis task, obtain classroom transcript segments in real time.

[0061] The classroom question and answer analysis task is used to instruct the smart device to start the question and answer analysis of the current class. The classroom question and answer analysis task can be actively triggered by the user, that is, the classroom question and answer analysis task is triggered according to the user's active sending of instructions.

[0062] In some possible implementations, the classroom question and answer analysis task can be automatically triggered at a scheduled time. For example, the classroom question and answer analysis task can be automatically triggered at a scheduled time according to the class start time.

[0063] In some possible implementations, the classroom question and answer analysis task can be automatically triggered by the smart device based on the keywords monitored in the classroom. For example, when questions and answers appear randomly in the classroom, when the smart device monitors the keyword "class", the classroom question and answer analysis task is automatically triggered. Or, there is a fixed question and answer session in the classroom, and when the smart device monitors the keyword "enter the question and answer session", the classroom question and answer analysis task is automatically triggered.

[0064] The traditional classroom question and answer analysis method needs to wait for the class to end and perform classroom question and answer analysis based on the complete classroom record. In the embodiment of the present application, based on the classroom question and answer analysis task, the classroom record segmentation is obtained in real time, and the classroom question and answer analysis task can be performed for the classroom record segmentation, without waiting for the class to end and obtain the complete classroom record before starting, which can effectively improve the efficiency of question and answer analysis.

[0065] In this embodiment, there are several classroom record segments, and several classroom record segments obtained based on the same classroom question and answer analysis task are spliced ​​to form a complete classroom record of the classroom question and answer analysis task.

[0066] In some possible implementations, a transcript segment includes at least one round of conversation between a teacher and a student.

[0067] In some possible implementations, the total class duration is fixed. At the beginning of the class, based on the classroom question and answer analysis task, the live recording of the classroom video or audio is started. When the total duration of the live recording reaches the total class duration, the class is determined to be over, and the live recording of the class corresponding to this classroom question and answer analysis task is ended.

[0068] In some possible implementations, the classroom includes clear technical signs, such as the teacher's words announcing the end of get out of class or the ringing of a bell, and based on the classroom question and answer analysis task, the live recording of the classroom video or audio is started. When the teacher's words announcing the end of class or the ringing of a bell are detected, the class is determined to be over, and the live recording of the classroom corresponding to the classroom question and answer analysis task is completed.

[0069] In some possible implementations, in order to ensure the effectiveness of question-and-answer analysis, the duration of the classroom recording segment must reach a preset recording duration threshold to avoid invalid analysis due to the short duration of the recording segment. In actual scenarios, after starting the live recording of the classroom video or audio based on the classroom question-and-answer analysis task, when the live recording reaches the preset recording duration threshold, a classroom recording segment can be obtained, the current time is reset, and the live recording is restarted. When the live recording reaches the preset recording duration threshold again, the next classroom recording segment can be obtained, and so on, until the end of the class, the acquisition of classroom recording segments is completed.

[0070] The preset recording duration threshold can be determined according to the total duration of the class. For example, if a class lasts 45 minutes, the preset recording duration threshold can be 15 minutes. When the total recording of the class reaches the preset recording duration threshold, a class recording segment can be obtained. In this embodiment, the duration of each class recording segment is the same and reaches the preset recording duration threshold.

[0071] It should be noted that if the live recording duration does not reach the preset recording duration threshold when the time is re-counted to the end of the class, the live recording that started cumulatively will also be determined as a class recording segment, which is the last class recording segment of this class question and answer analysis task. That is, the duration of the non-last class recording segment reaches the preset recording duration threshold, while the duration of the last class recording segment is less than or equal to the preset recording duration threshold.

[0072] In some possible implementations, in order to ensure the effectiveness of question and answer analysis, the number of characters in the classroom record segment must reach a preset record character number threshold to avoid invalid analysis due to too few characters in the record segment. In actual scenarios, based on the classroom question and answer analysis task, after starting the live recording of the classroom video or audio, the live recording is converted into text in real time, and the number of characters in the text is accumulated. When the accumulated number of characters reaches the preset record character number threshold, a classroom record segment can be obtained, and the text content corresponding to the classroom record segment is the record text segment; the character number record is cleared, and the number of characters in the text converted from the new live recording is accumulated again. When the accumulated number of characters reaches the preset record character number threshold again, a classroom record segment can be obtained, and so on, until the end of the class, the acquisition of the classroom record segment ends. The preset record character number threshold can be determined based on the number of characters in the historical record text segment corresponding to the historical classroom record segment.

[0073] It should be noted that if the number of characters is accumulated again at the end of the class, and the number of characters does not reach the preset threshold of the number of characters in the recording, the live recording of this time will also be determined as a class recording segment, and the class recording segment is the last class recording segment of this class question and answer analysis task, and the text corresponding to the last class recording segment is the last recording text segment. That is, the number of characters in the non-last class recording segment reaches the preset threshold of the number of characters in the recording, and the number of characters in the last class recording segment is less than or equal to the preset threshold of the number of characters in the recording.

[0074] Step S102: Convert each of the classroom transcript segments acquired in real time into transcript text segments and input them into the question-answer analysis model.

[0075] The classroom record can be a video or an audio. In this embodiment, each classroom record segment obtained in real time is converted into a record text segment and then input into the question-answer analysis model for processing, without waiting for the complete classroom record, which can effectively improve the completion efficiency of the classroom question-answer analysis task.

[0076] In some possible implementations, the transcript text segments are segmented according to a preset sequence length to obtain a plurality of sequences. The plurality of sequences are input into the question-and-answer analysis model in sequence. The preset sequence length can be determined based on business scenarios and experiments, thereby optimizing the performance of the model on the question-and-answer analysis task on the input side. Specifically, by segmenting the transcript text into a plurality of sequences and then inputting them into the question-and-answer analysis model, it is easier for the model to understand and analyze, thereby further improving the performance and efficiency of the model on the question-and-answer analysis task.

[0077] As a possible implementation of this application, Figure 2 A specific implementation process of sequentially inputting the plurality of sequences into the question-and-answer analysis model in the classroom question-and-answer analysis method provided in an embodiment of the present application is shown, and is described in detail as follows:

[0078] A1: If the sequence length of the end sequence in the transcript text segment is less than the preset sequence length, it is determined whether the transcript text segment is the end segment. The end class transcript segment has an end segment mark. If the transcript text segment has an end segment mark, the transcript text segment is the end segment.

[0079] A2: If the transcript text segment is the end segment, the end sequence of the end segment is input into the question-answer analysis model.

[0080] A3: If the recorded text segment is not the end segment, the end sequence of the recorded text segment is suspended, and the next recorded text segment is monitored. Suspending the end sequence of the recorded text segment means pausing the processing of the end sequence of the recorded text segment.

[0081] A4: merging the end sequence of the transcript text segment with the next transcript text segment detected, and segmenting them according to the preset sequence length to obtain a plurality of new sequences.

[0082] In this embodiment, when the sequence length of the end sequence in the transcript text segment is less than the preset sequence length, and the transcript text segment is not the end segment, the processing pointer is stopped at the beginning of the end sequence of the transcript text segment, and the next transcript text segment is awaited. When the next transcript text segment is detected, the paused end sequence is spliced ​​with the next transcript text segment, and the spliced ​​text is segmented according to the preset sequence length to obtain several new sequences, which can avoid semantic truncation, thereby retaining more complete context information, and is conducive to the large model to perform question-answering analysis tasks.

[0083] A5: Input the new sequences into the question-answer analysis model in sequence.

[0084] In the embodiment of the present application, for the end sequence in a transcript text segment, if its length is less than the preset sequence length and the corresponding transcript text segment is indeed the end segment, directly inputting the end sequence of the transcript text segment into the model can ensure that all information is processed in a timely manner, avoiding delayed processing due to waiting for more data. If the end sequence is not in the end segment, but the sequence length is insufficient due to the temporary state of text segmentation, it is suspended and waited for the next transcript text segment to be merged and processed, which can ensure the integrity of the data and avoid missed judgments or misjudgments due to improper data segmentation.

[0085] The large question-and-answer analysis model in the embodiment of the present application is used for classroom question-and-answer analysis, and is obtained by fine-tuning the large language model using a training set that meets the custom specifications. The large question-and-answer analysis model can efficiently understand and analyze questions and answers in the classroom through natural language processing and deep learning technology, assist teaching staff to better understand students' learning status, optimize teaching strategies, improve teaching quality, analyze students' thinking patterns and learning habits, and provide strong support for personalized teaching.

[0086] As a possible implementation of this application, Figure 3 A specific implementation process of obtaining a large question and answer analysis model in the classroom question and answer analysis method provided in an embodiment of the present application is shown, and is described in detail as follows:

[0087] Step S301: construct a training set that meets custom specifications.

[0088] Collect and compile a large number of classroom recording samples and their labels in the target domain (education field). The classroom recordings include dialogue samples between teachers and students. The collected classroom recordings cover different teaching scenarios, subjects and age groups, which can ensure the diversity of training samples.

[0089] The custom specification includes the specification of the sample label. The format of the sample label is structured and processed to obtain the sample label of the classroom record structure. In this embodiment, Figure 3.1 As shown, the structured sample labels involved are presented in the form of a dictionary and stored in a computer storage medium in JSON format. Specifically, each key-value pair in the dictionary corresponds to each element in the question and answer analysis. For example, the question and answer elements include questions and question types and answers and answer types. In the dictionary, there will be four sets of key values ​​corresponding to them, namely: question key-value pairs, question type key-value pairs, answer key-value pairs and answer type key-value pairs. In this embodiment, the structured labeling mode has strong scalability. With the evolution of the question and answer analysis task, the question and answer analysis elements may also change accordingly. It is only necessary to add or modify the key-value pairs in the dictionary and re-fine-tune the model to meet the needs. Using the structured sample labels in the embodiment of the present application to fine-tune the model can greatly improve the ability of the large model to extract various elements in the classroom question and answer analysis task. The specific indicators of model fine-tuning include: accuracy, recall rate, accuracy of question and answer interval positioning, and the degree to which the question and answer content is faithful to the original text.

[0090] In this embodiment, questions and answers in the classroom transcript can be manually annotated and organized in a specified structured format to generate structured sample labels.

[0091] In some possible implementations, specific prompt words and samples to be labeled are input into a large model with performance close to that of humans (such as GPT-4o and Qwen-Max), and structured sample labels including questions and their types and answers and their types are automatically generated. Using AI technology, automatic generation of sample labels can significantly shorten the model fine-tuning cycle.

[0092] In some possible implementations, the structured sample labels are serialized, and the serialized sample labels can be directly used for fine-tuning the large language model. For example, the sample labels in JSON format are serialized and converted into JSON Markdown strings. Examples of serialized sample labels are as follows: Figure 3.2 The abstract form of the serialized tag is: ```json\n{question and answer analysis results (JSON string)}\n```, where the content in {} is the JSON string after the sample tag in JSON format is serialized. ```json and ``` are Markdown symbols representing JSON.

[0093] The sample labels that have been structured and serialized are concatenated with the classroom transcripts to obtain a training sample. After performing the same operation on all collected classroom transcripts and their sample labels, a training set that meets the custom specifications is obtained.

[0094] Step S302: construct a fine-tuning task, where the fine-tuning task is used to instruct the large language model to learn to perform classroom question-and-answer analysis on the transcript text corresponding to the classroom transcript, and generate structured question-and-answer analysis elements.

[0095] The purpose of the fine-tuning task is to enable the large language model to learn to perform classroom question and answer analysis on the transcript text corresponding to the classroom transcript and generate structured question and answer analysis elements.

[0096] Step S303: Fine-tune the large language model based on the fine-tuning task and the training set until a large question-answering analysis model that completes the fine-tuning task is obtained.

[0097] In this embodiment, a pre-trained large language model is selected as a basis, such as Qwen, Llama or ChatGLM, etc. The large language model is fine-tuned based on the fine-tuning task and the training set that meets the custom specification. The specific process of fine-tuning training can refer to the prior art and will not be repeated here.

[0098] The embodiment of the present application utilizes specific fine-tuning tasks and a training set that meets custom planning to perform fine-tuning training on a large language model, which can improve the performance of the model's question-answering analysis.

[0099] Step S103: Use the question-and-answer analysis big model to perform classroom question-and-answer analysis on each of the transcript text segments until all transcript text segments corresponding to the classroom question-and-answer analysis task are processed, and then generate a big model question-and-answer analysis result.

[0100] In this embodiment, question and answer elements are identified and extracted from the transcript text segmentation, including questions and answers and their types. After the end segmentation of the transcript text in the classroom question and answer analysis task is processed, the number or proportion of questions and answers is counted, combined with all the extracted questions and their types and answers and their types, to generate a large model question and answer analysis result.

[0101] As a possible implementation of this application, Figure 4 A specific implementation process of step S103 in the classroom question and answer analysis method provided in an embodiment of the present application is shown, and is described in detail as follows:

[0102] B1: Extracting the question and answer elements in the transcript text segment, and determining whether the transcript text segment is the end segment.

[0103] B2: If the transcript text segment is not the end segment, the question and answer elements extracted from the segment are used as the intermediate analysis result.

[0104] B3: If the segmented record text is the last segment, then merge the Q&A elements extracted from the segment with all the intermediate analysis results to generate the large model Q&A analysis result corresponding to the classroom Q&A analysis task.

[0105] The embodiments of the present application flexibly process the segmented record text. Whether it is a non-last segment or a last segment, appropriate processing can be obtained. For non-last segments, Q&A elements can be extracted and saved as intermediate analysis results; for last segments, the Q&A elements extracted therefrom are merged with the previous intermediate analysis results, and the total number or proportion of questions and answers is statistically determined to generate the Q&A analysis result corresponding to the classroom Q&A analysis task, which is convenient for subsequent analysis, retrieval and display.

[0106] For long record texts, through segmented processing, not only can the processing efficiency be improved, but also each part can be processed and analyzed more carefully, reducing the error rate caused by processing the entire long record text at one time. By segmenting and extracting Q&A elements and merging them at the last segment, it can ensure that all Q&A information in the record text is completely retained, and avoid missed or misjudged caused by segmented processing.

[0107] As a possible implementation manner of the present application, Figure 5 shows a specific implementation process for optimizing the Q&A analysis result in the classroom Q&A analysis method provided by the embodiments of the present application, which is described in detail as follows:

[0108] C1: Obtain the record text corresponding to the classroom Q&A analysis task. This record text includes the text content of all classroom record segments obtained by the classroom Q&A analysis task.

[0109] C2: Use a rule-based classification algorithm to extract questions from the record text.

[0110] According to the characteristics of the questions, customize the rules for identifying questions. The rules can be expressed in the form of IF-THEN, that is, if certain conditions are met (such as containing specific words or sentence patterns), then it is considered a question.

[0111] Exemplarily, determine whether a sentence in the record text contains special words such as a question mark (?), "ma", and "what (at the end of the sentence)" to make a question judgment, and determine the sentence that meets the conditions as a question.

[0112] C3: Optimize the questions in the large model Q&A analysis result according to the questions.

[0113] The rule-based classification algorithm can accurately define and adjust the question extraction rules according to actual needs, so as to ensure the accuracy and pertinence of the extraction.

[0114] In an embodiment of the present application, the text similarity between the first question and the second question is calculated, wherein the first question includes the question in the large model question and answer analysis result, and the second question includes the question extracted from the transcript text by a rule-based classification algorithm; based on the text similarity, the questions in the large model question and answer analysis result are deduplicated and supplemented. If the text similarity between the first question and the second question is equal to or higher than the preset similarity threshold, the second question is removed; if the text similarity between the first question and the second question is lower than the preset similarity threshold, the second question is retained. The retained second question is merged into the large model question and answer analysis result generated by the large model to obtain an optimized large model question and answer analysis result.

[0115] As can be seen from the above, in the embodiment of the present application, the smart device obtains classroom transcript segments in real time based on the classroom question and answer analysis task, and converts each of the classroom transcript segments obtained in real time into transcript text segments and then inputs them into the question and answer analysis big model, and uses the question and answer analysis big model to perform classroom question and answer analysis on each of the transcript text segments until all the transcript text segments corresponding to the classroom question and answer analysis task are processed, and then the big model question and answer analysis results are automatically generated. There is no need to wait for the end of the class to analyze the complete classroom transcript, and there is no need to rely on manual record analysis. The time of classroom question and answer analysis can be effectively shortened, and the timeliness and accuracy of classroom question and answer analysis can be improved. At the same time, the question and answer analysis big model obtained by fine-tuning the big language model with a training set that meets the custom specifications in the present application solution generates a structured question and answer analysis result, which is conducive to improving the performance of the big model question and answer analysis and the granularity of accurate question and answer extraction.

[0116] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0117] Corresponding to the classroom question-answer analysis method described in the above embodiment, Figure 6 A structural block diagram of a classroom question and answer analysis device provided in an embodiment of the present application is shown. For ease of explanation, only the parts related to the embodiment of the present application are shown.

[0118] Reference Figure 6 The classroom question and answer analysis device includes: a transcript segment acquisition unit 61, a segment text input unit 62, and a question and answer analysis processing unit 63, wherein:

[0119] A real-time segmentation acquisition unit 61 is used to acquire real-time segmentation of classroom records based on the classroom question-answering analysis task;

[0120] Segmented text input unit 62, used to convert each of the classroom transcript segments obtained in real time into transcript text segments and then input them into the question-answer analysis model, wherein the question-answer analysis model is obtained by fine-tuning the large language model using a training set that meets the custom specification;

[0121] The question and answer analysis processing unit 63 is used to use the question and answer analysis big model to perform classroom question and answer analysis on each of the transcript text segments until all the transcript text segments corresponding to the classroom question and answer analysis task are processed, and then generate a big model question and answer analysis result.

[0122] As a possible implementation of the present application, the segmented text input unit 62 includes:

[0123] A text segmentation module is used to segment the transcript text into segments according to preset sequence lengths to obtain a plurality of sequences;

[0124] A text input module is used to input the plurality of sequences into the question-answer analysis model in sequence.

[0125] As a possible implementation of the present application, the above text input module is specifically used for:

[0126] If the sequence length of the end sequence in the transcript text segment is less than the preset sequence length, determining whether the transcript text segment is the end segment;

[0127] If the transcript text segment is the end segment, the end sequence of the end segment is input into the question-answer analysis model;

[0128] If the recorded text segment is not the end segment, suspending the end sequence of the recorded text segment and monitoring the next recorded text segment;

[0129] The end sequence of the recorded text segment is merged with the next recorded text segment detected, and divided according to the preset sequence length to obtain a plurality of new sequences;

[0130] The new sequences are input into the question-answer analysis model in sequence.

[0131] As a possible implementation of the present application, the question-answer analysis processing unit 63 includes:

[0132] An extraction and judgment module, used to extract the question and answer elements in the transcript text segment and judge whether the transcript text segment is the end segment;

[0133] An analysis and processing module, configured to use the question and answer elements extracted from the transcript text segment as an intermediate analysis result if the transcript text segment is not the end segment;

[0134] If the transcript text segment is the end segment, the question and answer elements extracted from the transcript text segment are merged with all the intermediate analysis results to generate a large model question and answer analysis result corresponding to the classroom question and answer analysis task.

[0135] As a possible implementation of the present application, the classroom question and answer analysis device further includes:

[0136] A transcript acquisition unit, used to acquire the transcript text corresponding to the classroom question and answer analysis task;

[0137] A question extraction unit, used to extract questions from the transcript text using a rule-based classification algorithm;

[0138] The question-and-answer analysis result optimization unit is used to optimize the questions in the large model question-and-answer analysis result according to the questions.

[0139] As a possible implementation of the present application, the question-answer analysis result optimization unit includes:

[0140] A similarity calculation module, used to calculate the text similarity between a first question and a second question, wherein the first question includes questions in the large model question-answering analysis result, and the second question includes questions extracted from the transcript text by a rule-based classification algorithm;

[0141] The analysis result optimization module is used to remove duplicates and supplement questions in the large model question and answer analysis results based on the text similarity.

[0142] As a possible implementation of the present application, the classroom question and answer analysis device further includes a model fine-tuning unit for:

[0143] Build a training set that meets custom specifications;

[0144] Constructing a fine-tuning task, wherein the fine-tuning task is used to instruct the large language model to learn to perform classroom question-and-answer analysis on the transcript text corresponding to the classroom transcript, and generate structured question-and-answer analysis elements;

[0145] Fine-tune the large language model based on the fine-tuning task and the training set until a large question-answering analysis model that completes the fine-tuning task is obtained.

[0146] As can be seen from the above, in the embodiment of the present application, the smart device obtains classroom transcript segments in real time based on the classroom question and answer analysis task, and converts each of the classroom transcript segments obtained in real time into transcript text segments and then inputs them into the question and answer analysis big model, and uses the question and answer analysis big model to perform classroom question and answer analysis on each of the transcript text segments until all the transcript text segments corresponding to the classroom question and answer analysis task are processed, and then the big model question and answer analysis results are automatically generated. There is no need to wait for the end of the class to analyze the complete classroom transcript, and there is no need to rely on manual record analysis. The time of classroom question and answer analysis can be effectively shortened, and the timeliness and accuracy of classroom question and answer analysis can be improved. At the same time, the question and answer analysis big model obtained by fine-tuning the big language model with a training set that meets the custom specifications in the present application solution generates a structured question and answer analysis result, which is conducive to improving the performance of the big model question and answer analysis and the granularity of accurate question and answer extraction.

[0147] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0148] The present application also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, Figures 1 to 5 The steps of any classroom question and answer analysis method are represented.

[0149] The embodiment of the present application also provides an intelligent device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, Figures 1 to 5 The steps of any classroom question and answer analysis method are represented.

[0150] The embodiment of the present application also provides a computer program product, when the computer program product is run on a smart device, the smart device executes the following Figures 1 to 5 The steps of any classroom question and answer analysis method are represented.

[0151] Figure 7 Schematic diagram of a smart device provided by an embodiment of the present application. Figure 7 As shown, the smart device 7 of this embodiment includes: a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and executable on the processor 70. When the processor 70 executes the computer program 72, the steps in the above-mentioned classroom question-answer analysis method embodiments are implemented, for example Figure 1Alternatively, when the processor 70 executes the computer program 72, the functions of each module / unit in the above-mentioned device embodiments are realized, for example Figure 6 The functions of units 61 to 63 are shown.

[0152] Exemplarily, the computer program 72 may be divided into one or more modules / units, which are stored in the memory 71 and executed by the processor 70 to complete the present application. The one or more modules / units may be a series of computer-readable instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 72 in the smart device 7.

[0153] The smart device 7 may include, but is not limited to, a processor 70 and a memory 71. Those skilled in the art will appreciate that Figure 7 It is only an example of the smart device 7 and does not constitute a limitation of the smart device 7. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the smart device 7 may also include input and output devices, network access devices, buses, etc.

[0154] The processor 70 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0155] The memory 71 may be an internal storage unit of the smart device 7, such as a hard disk or memory of the smart device 7. The memory 71 may also be an external storage device of the smart device 7, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the smart device 7. Further, the memory 71 may also include both an internal storage unit and an external storage device of the smart device 7. The memory 71 is used to store the computer program and other programs and data required by the smart device. The memory 71 may also be used to temporarily store data that has been output or is to be output.

[0156] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0157] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0158] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the device / intelligent device, a recording medium, a computer memory, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), an electric carrier signal, a telecommunication signal, and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0159] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0160] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A classroom question-answer analysis method, characterized in that: include: Based on the classroom question and answer analysis task, the classroom transcript segmentation is obtained in real time; Each of the classroom transcript segments obtained in real time is converted into transcript text segments and then input into a large question-answer analysis model, wherein the large question-answer analysis model is obtained by fine-tuning a large language model using a training set that meets custom specifications; The question-and-answer analysis big model is used to perform classroom question-and-answer analysis on each of the transcript text segments until all transcript text segments corresponding to the classroom question-and-answer analysis task are processed, thereby generating a big model question-and-answer analysis result.

2. The classroom question-answer analysis method according to claim 1, characterized in that: The step of converting each of the classroom transcript segments obtained in real time into transcript text segments and then inputting them into the question-answer analysis model comprises: The recorded text is segmented according to a preset sequence length to obtain a plurality of sequences; The several sequences are input into the question-answer analysis model in sequence.

3. The classroom question-answer analysis method according to claim 2, characterized in that: The step of sequentially inputting the plurality of sequences into the question-answer analysis large model comprises: If the sequence length of the end sequence in the transcript text segment is less than the preset sequence length, determining whether the transcript text segment is the end segment; If the transcript text segment is the end segment, the end sequence of the end segment is input into the question-answer analysis model; If the recorded text segment is not the end segment, suspending the end sequence of the recorded text segment and monitoring the next recorded text segment; The end sequence of the recorded text segment is merged with the next recorded text segment detected, and divided according to the preset sequence length to obtain a plurality of new sequences; The new sequences are input into the question-answer analysis model in sequence.

4. The classroom question-answer analysis method according to claim 1, characterized in that: The step of using the question-and-answer analysis big model to perform classroom question-and-answer analysis on each of the transcript text segments until all transcript text segments corresponding to the classroom question-and-answer analysis task are processed, and then generating a big model question-and-answer analysis result comprises: Extracting question-answer elements and their attributes in the transcript text segment, and determining whether the transcript text segment is the end segment; If the transcript text segment is not the end segment, the question and answer elements extracted from the transcript text segment are used as the intermediate analysis result; If the transcript text segment is the end segment, the question and answer elements extracted from the transcript text segment are merged with all the intermediate analysis results to generate a large model question and answer analysis result corresponding to the classroom question and answer analysis task.

5. The classroom question-answer analysis method according to claim 1, characterized in that: After the step of generating the large model question-answering analysis result, the method further includes: Obtaining the transcript corresponding to the classroom question and answer analysis task; Using a rule-based classification algorithm to extract questions from the transcript text; The questions in the large model question and answer analysis results are optimized according to the questions.

6. The classroom question-answer analysis method according to claim 5, characterized in that: The step of optimizing the questions in the large model question-answering analysis result according to the questions includes: Calculating the text similarity between the first question and the second question, wherein the first question includes the question in the large model question-answering analysis result, and the second question includes the question extracted from the transcript text by a rule-based classification algorithm; Based on the text similarity, the questions in the large model question and answer analysis results are deduplicated and supplemented.

7. According to the classroom question and answer analysis method according to any one of claims 1 to 6, the step of fine-tuning the large language model using the training set that meets the custom specification to obtain the large model for question and answer analysis comprises: Build a training set that meets custom specifications; Constructing a fine-tuning task, wherein the fine-tuning task is used to instruct the large language model to learn to perform classroom question-and-answer analysis on the transcript text corresponding to the classroom transcript, and generate structured question-and-answer analysis elements; Fine-tune the large language model based on the fine-tuning task and the training set until a large question-answering analysis model that completes the fine-tuning task is obtained.

8. A classroom question-answer analysis device, characterized in that: include: A real-time segment acquisition unit, used to acquire real-time segmentation of classroom records based on classroom question-answer analysis tasks; A segmented text input unit is used to convert each of the classroom transcript segments obtained in real time into transcript text segments and then input them into the question-answer analysis model, wherein the question-answer analysis model is obtained by fine-tuning the large language model using a training set that meets the custom specification; The question and answer analysis processing unit is used to use the question and answer analysis big model to perform classroom question and answer analysis on each of the transcript text segments until all the transcript text segments corresponding to the classroom question and answer analysis task are processed, and then generate a big model question and answer analysis result.

9. An intelligent device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the classroom question and answer analysis method as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the classroom question and answer analysis method as described in any one of claims 1 to 7 is implemented.

Images