Teaching behavior information extraction and processing method and equipment based on content masking

Through content masking processing and text embedding technology, combined with large language models, teaching behavior information is extracted, and the high cost and low efficiency problems of teaching behavior information extraction and processing in the existing technology are solved, and automated extraction and rich teaching behavior data records are realized. .

CN119088928APending Publication Date: 2024-12-06SHANGHAI NORMAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411190544.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

It is difficult for the existing technology to efficiently extract and process teaching behavior information, especially after the development of large-scale model technology, the conversion and processing of unstructured data is high, and there is a lack of automated tools to form structured data.

Method used

By obtaining teacher teaching text data and classroom content data, slicing processing and naming entity recognition are carried out, content masking processing and text embedding technology are used, and teaching behavior information is extracted to form a structured behavior label and knowledge base using clustering and large language models. .

Benefits of technology

It realizes the automated extraction and identification of teaching behavior information, reduces the cost of manual sorting, improves data processing efficiency, and enriches the classification records of teaching behavior data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119088928A_ABST
    Figure CN119088928A_ABST
Patent Text Reader

Abstract

The invention relates to a teaching behavior information extraction and processing method and equipment based on content masking, and the method comprises the steps: obtaining text data corresponding to the teaching voice of a teacher and teaching classroom content data, and carrying out the slicing processing, thereby obtaining a teaching language text slice and a classroom content slice, performing named entity recognition on the teaching classroom content data to obtain a teaching content entity set; for the teaching language text slices which are not matched with the classroom content slices, matching and labeling entities in the teaching language text slices by utilizing a teaching content entity set to realize content masking processing; performing embedding processing on the teaching language text slices subjected to content masking processing to obtain text vectors, clustering the plurality of text vectors, summarizing and removing miscellaneous classes by utilizing a large language model subjected to data prompt in advance based on the clustered text vectors corresponding to each class label to obtain text vectors with behavior labels as classifications, and performing classification processing on the text vectors to obtain the teaching language text slices subjected to content masking processing. And teaching behavior information extraction is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and device for extracting and processing teaching behavior information based on content masking. Background Art

[0002] Before the development of artificial intelligence, the collection and analysis of teachers' teaching behaviors were mostly manually labeled. For example, the commonly used classroom teaching behavior analysis methods - student and teacher classroom behavior analysis method (ST classroom analysis method) and Flanders Interaction Analysis System (FIAS) use manual coding and timing to collect data. Although this method has high accuracy, it costs a huge amount of money.

[0003] At present, with the rapid development of big model technology, the importance of data is becoming increasingly prominent. In particular, the value of unstructured data, such as text, images, videos, etc., in big model training cannot be ignored. However, converting these unstructured data into structured data involves not only ensuring data quality, but also cost control, which has become an important challenge.

[0004] Teaching behavior refers to the teaching methods of teachers during the teaching process, which include the way of explaining knowledge, instructing students, and other teaching behaviors. Taking Chinese class as an example, it can include: learning guidance, guidance, and explanation of text content.

[0005] Currently, artificial intelligence has demonstrated powerful data processing capabilities in many fields. In particular, in natural language processing and computer vision, advanced algorithms and models make it more feasible to extract and analyze key information in video data. Through automated tools, the explanation content and interactive behaviors in teaching videos can be identified, but the teaching behaviors are not processed to form structured data.

[0006] In summary, how to achieve efficient extraction and processing of teaching behavior information is an urgent problem to be solved. Summary of the invention

[0007] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a teaching behavior information extraction and processing method and device based on content masking, so as to realize the efficient extraction and processing of teaching behavior information.

[0008] The purpose of the present invention can be achieved by the following technical solutions:

[0009] One aspect of the present invention provides a method for extracting and processing teaching behavior information based on content masking, comprising the following steps:

[0010] Step S1, obtaining the teacher's teaching text data and the classroom content data, obtaining the teaching language text slices and the classroom content slices by slicing them respectively, and obtaining the teaching content entity set by performing named entity recognition on the classroom content data;

[0011] Step S2, for the teaching language text slices that do not match the classroom content slices, use the teaching content entity set to match and annotate entities in the teaching language text slices to achieve content masking processing;

[0012] Step S3, embedding processing is performed on the teaching language text slices after content masking processing to obtain text vectors, multiple text vectors are clustered, and based on the text vectors corresponding to each class label after clustering, a large language model that has been pre-data-prompted is used to summarize and eliminate miscellaneous categories to obtain text vectors classified by behavior labels, thereby realizing the extraction of teaching behavior information.

[0013] As a preferred technical solution, in step S1, the slicing process includes:

[0014] Step S101, convert the teacher's teaching video and / or recording to obtain the teacher's teaching text data, segment the teacher's teaching text data according to language pauses, obtain teaching language text slices, and divide the classroom content into short sentences according to symbols.

[0015] As a preferred technical solution, in the step S1, the process of obtaining a teaching content entity set by performing named entity recognition on the teaching classroom content data includes:

[0016] Step S102, obtaining a named entity recognition model, wherein the named entity recognition model is obtained based on classroom content entity training;

[0017] Step S103, use the named entity recognition model to perform named entity recognition on the classroom content slices, extract the content entities of the current classroom and label the content entities to obtain a teaching content entity set.

[0018] As a preferred technical solution, the label of the content entity includes at least one of teaching content characters, teaching content scenes, teaching content events, and teaching content items.

[0019] As a preferred technical solution, the step S2 includes:

[0020] Determine whether the teaching language text slice and the classroom content slice match. If so, mark the teaching language text slice as teaching content. If not, use the teaching content entity set to match and mark the entities in the teaching language text slice to achieve content masking processing.

[0021] As a preferred technical solution, in step S2, the process of matching and marking entities in the teaching language text slice using the teaching content entity set includes:

[0022] The teaching content entity set is used to match entities appearing in the teaching language text slice, and the existing entities are replaced according to entity category labels.

[0023] As a preferred technical solution, the step S3 includes:

[0024] Step S301, using a text embedding and training model trained based on educational text corpus, vectorizing the teaching language text slice after content masking to form a text vector that can represent the teaching language text slice after content masking;

[0025] Step S302, performing unsupervised classification on the multiple text vectors by using a clustering algorithm to form multiple class labels;

[0026] Step S303, using the large language model prompted by the teaching method data, summarize the clustered class labels and eliminate the miscellaneous ones to obtain the text vector classified by the behavior label.

[0027] As a preferred technical solution, it also includes:

[0028] Step S4, constructing a teaching behavior knowledge base based on the text vectors classified by behavior labels.

[0029] Another aspect of the present invention provides an electronic device, comprising: one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the aforementioned content-masking-based teaching behavior information extraction and processing method.

[0030] Another aspect of the present invention provides a computer-readable storage medium comprising one or more programs for execution by one or more processors of an electronic device, wherein the one or more programs include instructions for executing the aforementioned content-masking-based teaching behavior information extraction and processing method.

[0031] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0032] (1) Through the text processing of teaching videos or teaching voices by machine learning and natural language processing algorithms in deep learning, the automatic extraction and recognition of teaching behavior information in teacher teaching records in video or voice form is realized.

[0033] (2) Through the technical implementation of masking the classroom content on the language text, the weight of the teaching content in the vector representation is reduced, and the influence of the content on the clustering is reduced during clustering, thereby improving the effect of identifying and extracting teaching behavior information.

[0034] (3) Through the extraction and processing of teaching behavior information, different teaching behaviors can be classified and recorded to form a teaching behavior knowledge base, enriching the data on teaching behavior in the current education field. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flowchart of the teaching behavior information extraction and processing method based on content masking in the embodiment. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0037] Example 1

[0038] In view of the problems existing in the above-mentioned prior art, this embodiment provides a teaching behavior information extraction and processing method based on content masking, which aims to realize the automatic processing of teaching video data through artificial intelligence methods and convert it into structured behavior and content data forms. This method can not only improve the efficiency of data processing, but also significantly reduce the cost of manual data sorting while ensuring data quality.

[0039] See also Figure 1 , the method comprises the following steps:

[0040] Step S1, obtain the text data corresponding to the teacher's teaching voice and the classroom content data, obtain the teaching language text slices and classroom content slices by slicing them respectively, and obtain the teaching content entity set by performing named entity recognition on the classroom content data. Among them, the teacher's teaching text data includes excellent teacher teaching records, which are in the form of video and / or recording. The classroom content data includes one or more forms of knowledge points, texts, and teaching objectives.

[0041] This step mainly includes:

[0042] Step S101, slicing processing: convert the teacher's teaching video and / or recording to obtain the teacher's teaching text data, segment the teacher's teaching text data according to language pauses, obtain teaching language text slices, and divide the classroom content into short sentences according to symbols.

[0043] Step S102, obtaining a named entity recognition model, where the named entity recognition model is obtained based on classroom content entity training;

[0044] Step S103, use the named entity recognition model to perform named entity recognition on the classroom content slices, extract the content entities of the current classroom and label the content entities to obtain a set of teaching content entities. Specifically, the labels of the content entities include teaching content characters, teaching content scenes, teaching content events and teaching content items. The content entities include different knowledge labels for different subjects.

[0045] Step S2, content masking link: for the teaching language text slices that do not match the classroom content slices, use the teaching content entity set to match and mark the entities in the teaching language text slices to achieve content masking processing.

[0046] Specifically, first, determine whether the teaching language text slice and the classroom content slice match. If so, mark the teaching language text slice as teaching content. If not, use the teaching content entity set to match and mark the entities in the teaching language text slice to achieve content masking. Then, use the teaching content entity set to match the entities that appear in the teaching language text slice, and replace the existing entities according to the entity category label.

[0047] Step S3, text embedding is performed on the teaching language text slices after content masking to obtain text vectors, multiple text vectors are clustered, and for each class label after clustering, a large language model that has been pre-data-prompted is used to summarize and eliminate miscellaneous categories to obtain text vectors classified by behavior labels, thereby realizing the extraction of teaching behavior information.

[0048] This step specifically includes:

[0049] Step S301, using the text embedding and training model trained based on the educational text corpus, vectorize the teaching language text slices after content masking to form a text vector that can represent the teaching language text slices after content masking.

[0050] Step S302: performing unsupervised classification on multiple text vectors through a clustering algorithm to form multiple class labels. Specifically, the class labels may be 0, 1, or 2.

[0051] Step S303, using the large language model prompted by the teaching method data, summarize the clustered class labels and eliminate the miscellaneous ones to obtain text vectors classified by behavior labels. By summarizing the text vectors corresponding to the class labels through the large model, the class labels are actually labeled according to the knowledge in the field of education.

[0052] Step S4, constructing a teaching behavior knowledge base based on the text vectors classified by the behavior labels. Specifically, the teaching behaviors that have been classified and automatically cleaned are stored in a teaching behavior knowledge base according to different categories.

[0053] The constructed teaching behavior knowledge base stores the relationship between vectorized teaching language text fragments and teaching behavior types. The extracted teaching behavior information is combined with the teaching content associated with the teaching behavior information, which means that this teaching behavior can be used to teach the teaching content. By using the teaching record data of excellent teachers, the knowledge base formed in this way can use the knowledge vector library of the lesson plan design model as data support.

[0054] In summary, this method realizes the automatic label extraction of teaching behaviors in teaching records, uses the method of teaching content masking to improve the behavior extraction effect, and finally constructs a teaching behavior knowledge base of different categories. This method uses the machine learning and natural language processing algorithms in deep learning to process teaching videos or teaching voices from the text, and realizes the automatic extraction and recognition of teaching behavior information in the teacher's teaching records in the form of video or voice. The technical implementation of masking the classroom content on the language text improves the effect of identifying and extracting teaching behaviors. It realizes the classification and recording of different teaching behaviors to form a teaching behavior knowledge base, enriching the current teaching behavior data in the field of education.

[0055] Example 2

[0056] This embodiment provides an electronic device, including: one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the teaching behavior information extraction and processing method based on content masking as described in Example 1.

[0057] Example 3

[0058] This embodiment provides a computer-readable storage medium, including one or more programs for execution by one or more processors of an electronic device, wherein the one or more programs include instructions for executing the teaching behavior information extraction and processing method based on content masking as described in Example 1.

[0059] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for extracting and processing teaching behavior information based on content masking, characterized in that: The steps include: Step S1, obtaining text data corresponding to the teacher's teaching voice and classroom content data, obtaining teaching language text slices and classroom content slices by slicing them respectively, and obtaining a teaching content entity set by performing named entity recognition on the classroom content data; Step S2, for the teaching language text slices that do not match the classroom content slices, use the teaching content entity set to match and annotate entities in the teaching language text slices to achieve content masking processing; Step S3, embedding processing is performed on the teaching language text slices after content masking processing to obtain text vectors, multiple text vectors are clustered, and based on the text vectors corresponding to each class label after clustering, a large language model that has been pre-data-prompted is used to summarize and eliminate miscellaneous categories to obtain text vectors classified by behavior labels, thereby realizing the extraction of teaching behavior information.

2. According to the teaching behavior information extraction and processing method based on content masking according to claim 1, it is characterized in that: In step S1, the slicing process includes: Step S101, convert the teacher's teaching video and / or recording to obtain the teacher's teaching text data, segment the teacher's teaching text data according to language pauses, obtain teaching language text slices, and divide the classroom content into short sentences according to symbols.

3. The method for extracting and processing teaching behavior information based on content masking according to claim 1 is characterized in that: In the step S1, the process of obtaining a teaching content entity set by performing named entity recognition on the teaching classroom content data includes: Step S102, obtaining a named entity recognition model, wherein the named entity recognition model is obtained based on classroom content entity training; Step S103, use the named entity recognition model to perform named entity recognition on the classroom content slices, extract the content entities of the current classroom and label the content entities to obtain a teaching content entity set.

4. The method for extracting and processing teaching behavior information based on content masking according to claim 3 is characterized in that: The label of the content entity includes at least one of teaching content characters, teaching content scenes, teaching content events, and teaching content items.

5. The method for extracting and processing teaching behavior information based on content masking according to claim 1 is characterized in that: The step S2 comprises: Determine whether the teaching language text slice and the classroom content slice match. If so, mark the teaching language text slice as teaching content. If not, use the teaching content entity set to match and mark the entities in the teaching language text slice to achieve content masking processing.

6. The method for extracting and processing teaching behavior information based on content masking according to claim 1, characterized in that: In the step S2, the process of matching and labeling entities in the teaching language text slice using the teaching content entity set includes: The teaching content entity set is used to match entities appearing in the teaching language text slice, and the existing entities are replaced according to entity category labels.

7. The method for extracting and processing teaching behavior information based on content masking according to claim 1 is characterized in that: The step S3 comprises: Step S301, using a text embedding and training model trained based on educational text corpus, vectorizing the teaching language text slice after content masking to form a text vector that can represent the teaching language text slice after content masking; Step S302, performing unsupervised classification on the multiple text vectors by using a clustering algorithm to form multiple class labels; Step S303, using the large language model prompted by the teaching method data, summarize the clustered class labels and eliminate the miscellaneous ones to obtain the text vector classified by the behavior label.

8. The method for extracting and processing teaching behavior information based on content masking according to claim 1 is characterized in that: Also includes: Step S4, constructing a teaching behavior knowledge base based on the text vectors classified by behavior labels.

9. An electronic device, characterized in that: include: One or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the teaching behavior information extraction and processing method based on content masking as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: It comprises one or more programs executed by one or more processors of an electronic device, wherein the one or more programs comprise instructions for executing the teaching behavior information extraction and processing method based on content masking as described in any one of claims 1-8.