CN-121982695-A - Micro class extraction method, related method and related device

CN121982695ACN 121982695 ACN121982695 ACN 121982695ACN-121982695-A

Abstract

The invention discloses a micro-class extraction method, a related method and a related device, and relates to the technical field of resource fragmentation, wherein the micro-class extraction method comprises the steps of obtaining a target course video of a target subject; the method comprises the steps of analyzing the content of a target course video to generate target text data with a time stamp, slicing the target text data based on a large slicing model to obtain a slicing result at least comprising the start and stop time of a knowledge point slice, training a universal large model base by adopting training text data marked with a real knowledge point slicing result and aided with thinking process data for obtaining the real knowledge point slicing result, training a universal large model base to obtain the training text data, wherein the training text data is text data corresponding to a course video sample of a target subject, and extracting the knowledge point video slice from the target course video according to the start and stop time of the knowledge point slice to obtain a knowledge point video micro-class. The micro-class extraction method disclosed by the invention can automatically extract accurate knowledge point video micro-classes from the course video.

Inventors

ZHA HONGYU
WANG YONGHAI
WANG YU
WANG YING
LI YONGBIN

Assignees

科大讯飞股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260409

Claims (18)

1. A micro lesson extraction method, comprising: acquiring a target course video of a target subject; analyzing the content of the target course video to generate corresponding target text data with a time stamp; Carrying out knowledge point slicing on the target text data based on a large slicing model to obtain knowledge point slicing results, wherein the knowledge point slicing results at least comprise start and stop time of knowledge point fragments, the large slicing model adopts training text data marked with real knowledge point slicing results and is supplemented with thinking process data for obtaining the real knowledge point slicing results, and a universal large model base is trained to obtain the training text data which is text data with time stamps and corresponds to course video samples of the target subjects; and extracting knowledge point video clips from the target course video according to the start-stop time of the knowledge point clips to obtain knowledge point video micro-courses.
2. The micro lesson extraction method of claim 1, wherein the parsing the content of the target lesson video to generate corresponding time-stamped target text data comprises: performing audio and video separation on the target course video to obtain audio data and silent video data; and processing the audio data into standard, correct and structured text data with time stamps and speaker labels by combining the video content of the silent video data to obtain target text data corresponding to the target course video.
3. The micro-class extraction method of claim 2, wherein said processing the audio data into canonical, correct, time stamped and speaker tagged structured text data in conjunction with the video content of the silent video data, comprises: performing voice recognition on the audio data to obtain a recognition text with a time stamp, and distinguishing the teacher utterance from the student utterance by the recognition text to obtain a text with the time stamp and a speaker tag; extracting a key video frame from the silent video data, and extracting text information from the key video frame; Based on a large language model, combining text information extracted from the key video frames, correcting semantic, grammar and key information of the text with the timestamp and the speaker tag, and performing long sentence segmentation processing, formula standardization processing and semantic-based paragraph division processing on the corrected text to obtain a structured text with the standard, correct, timestamp and the speaker tag.
4. The micro lesson extraction method according to claim 1, wherein the slicing the knowledge point based on the sliced large model to obtain knowledge point slicing results includes: based on a large slicing model, combining the knowledge graph of the target subject and the course standard, slicing the knowledge points of the target text data to obtain knowledge point slicing results; or based on a large slicing model, combining a knowledge point list of chapters to which the target course video belongs, and slicing the knowledge points of the target text data to obtain knowledge point slicing results.
5. The micro lesson extraction method according to claim 1, wherein the slicing the knowledge point based on the sliced large model to obtain knowledge point slicing results includes: Based on a large slicing model, carrying out preliminary knowledge point slicing on the target text data to obtain a preliminary knowledge point slicing result; And optimizing the preliminary knowledge point segmentation result according to the knowledge graph and the course standard of the target subject and knowledge related to the knowledge point slicing based on the slice large model to obtain a final knowledge point segmentation result.
6. The micro lesson extraction method according to claim 1, wherein the slicing the knowledge point based on the sliced large model to obtain knowledge point slicing results includes: Acquiring a course teaching outline of the target subject; Slicing the course teaching outline according to the teaching links to obtain a teaching outline segmentation result; according to the teaching outline segmentation result, filtering text fragments which are irrelevant to knowledge point slicing in the target text data; And based on the large slicing model, slicing the knowledge points of the filtered text data to obtain knowledge point slicing results.
7. The method according to claim 1, wherein the knowledge point segmentation result further comprises a knowledge point micro-class title, a knowledge point micro-class abstract, and text content of the knowledge point micro-class; the micro class extraction method further comprises the following steps: Constructing metadata comprising the knowledge point micro-class title, the knowledge point micro-class abstract and a knowledge point label; packaging and encapsulating the metadata and the text content of the knowledge point micro lessons with the knowledge point video micro lessons to obtain complete knowledge point video micro lesson resources; And associating the knowledge point video micro-class resource with a corresponding knowledge point in the knowledge graph of the target subject.
8. The micro lesson extraction method of claim 1, further comprising: based on the large slicing model, combining with typical example problem slicing rules, carrying out typical example problem slicing on the target text data to obtain a slicing result at least comprising the start and stop time of typical example problem fragments, and extracting typical example problem video fragments from the target course video according to the start and stop time of the typical example problem fragments to obtain typical example problem video micro-class; And/or based on the large slice model, combining a scene material slicing rule, slicing the scene material data to obtain a slicing result at least comprising the start and stop time of the scene material fragments, and extracting scene material video fragments from the target course video according to the start and stop time of the scene material fragments to obtain scene material video micro-lessons; and/or based on the large slice model and combining with a classroom activity slicing rule, slicing the target text data to obtain a slicing result at least comprising the start and stop time of the classroom activity segment, and extracting a classroom activity video segment from the target course video according to the start and stop time of the classroom activity segment to obtain the classroom activity video micro-class.
9. The micro lesson extraction method of claim 8, wherein the performing a typical topic slicing on the target text data based on the sliced large model in combination with a typical topic slicing rule comprises: Based on the large-scale model, combining typical example question slicing rules, the keywords and the typical example question presentation modes in the class, carrying out typical example question slicing on the target text data; based on the large slice model, in combination with a scene material slicing rule, the method for slicing the scene material of the target text data comprises the following steps: based on the large slice model, combining the scene type and a slice rule corresponding to the scene type in a scene material slice rule, and slicing the scene material of the target text data; based on the large slice model, the classroom activity slicing is performed on the target text data by combining with a classroom activity slicing rule, and the method comprises the following steps: And based on the large slicing model, combining the activity type and a slicing rule corresponding to the activity type in the class activity slicing rules, and slicing the class activity of the target text data.
10. The micro lesson extraction method of claim 9, further comprising: Obtaining classical example problem solving process data according to the classical example problem fragments, and associating the classical example problem solving process data with the classical example problem video micro-class; Acquiring key elements and/or teaching intentions of the scene material according to the scene material fragments, and associating the key elements and/or teaching intentions with the scene material video micro-lessons; And acquiring an activity flow analysis result and/or a teaching target and/or teacher-student interaction information according to the classroom activity fragment, and associating the activity flow analysis result and/or the teaching target and/or the teacher-student interaction information with the classroom activity video micro-lesson, wherein the activity flow analysis result comprises part or all of activity organization form information, activity participation role information, activity process information and time allocation information of each link of an activity.
11. The micro lesson extraction method of claim 1, further comprising: acquiring a teaching plan document corresponding to the target course video; Carrying out content analysis on the teaching plan document to obtain a document analysis result; Identifying the document structure of the teaching plan document according to the document analysis result; obtaining structured teaching plan text data according to the document analysis result and the document structure, and slicing knowledge points of the structured teaching plan text data to obtain a knowledge point document; And associating the knowledge point document with a knowledge point video micro-class of the corresponding knowledge point, and/or associating the knowledge point document with the corresponding knowledge point in the knowledge graph of the target subject.
12. The micro lesson extraction method of claim 1, further comprising: Generating a teaching script of a target knowledge point, wherein the target knowledge point is a knowledge point corresponding to a knowledge point video micro-class; Generating teaching audio of the target knowledge points according to the teaching scripts of the target knowledge points to obtain knowledge point audio micro-lessons; And associating the knowledge point audio micro-lessons with the corresponding knowledge points in the knowledge graph of the target subject.
13. The micro lesson extraction method of claim 12, wherein the generating the teaching script of the target knowledge point comprises: Constructing a teaching script to generate a prompt word according to a target knowledge point and teaching design information of the target knowledge point, wherein the teaching design information comprises part or all of dialogue design logic, plot arrangement information, content organization mode information and character design and configuration information; And inputting the teaching script generation prompt word into a large language model to generate a teaching script, so as to obtain the teaching script of the target knowledge point.
14. A method for retrieving resources, comprising: Acquiring query information related to knowledge points input by a user; Retrieving resources matched with the query information from a pre-constructed resource library, wherein the resource library comprises resources associated with knowledge points in a subject knowledge graph, and the resources in the resource library comprise micro-class resources extracted by adopting the micro-class extraction method according to any one of claims 1-13.
15. An electronic device comprising at least one processor and a memory coupled to the processor, wherein: The memory is used for storing a computer program; the processor is configured to execute the computer program to cause the electronic device to implement the micro class extraction method according to any one of claims 1 to 13 or implement the resource retrieval method according to claim 14.
16. A computer storage medium carrying one or more computer programs which, when executed by an electronic device, enable the electronic device to implement the micro-class extraction method of any one of claims 1 to 13 or the resource retrieval method of claim 14.
17. A computer program product comprising computer readable instructions which, when run on an electronic device, cause the electronic device to implement a micro-class extraction method as claimed in any one of claims 1 to 13, or to implement a resource retrieval method as claimed in claim 14.
18. The resource processing management system is characterized by comprising a terminal and a server; the terminal is used for receiving the target course video uploaded by the user and uploading the target course video to the server; The server is configured to extract micro lessons from the target lesson video by using the micro lesson extraction method according to any one of claims 1 to 13.

Description

Micro class extraction method, related method and related device Technical Field The application relates to the technical field of resource fragmentation, in particular to a micro-class extraction method, a related method and a related device. Background The intelligent education public service platform gathers high-quality digital course resources, initially builds a resource supply system which is expanded transversely and longitudinally, and continuously promotes the improvement of the capabilities of intelligent analysis, personalized pushing, intelligent searching and the like. However, currently there are limitations in terms of flexibility in resource provisioning, whether it is the platform or the digital educational resource hierarchy described above. The resources converged by the existing platform are mostly systematic and structured complete courses, such as real-class records or whole-section teaching videos. Although the resources have advantages in the integrity of teaching contents, the granularity is large, the data structure is relatively solidified, and when a teacher carries out teaching or student fragmented learning on a single knowledge point, the teacher often has difficulty in directly stripping required knowledge point fragments from the complete videos. At present, the complete course resources are converted into knowledge point micro courses for scattered and fragmented use, and the knowledge point micro courses are usually highly dependent on manual screening or secondary recording of teachers, so that the cost of resource reconstruction is high, the efficiency is low, and the existing massive digital education resources cannot be fully utilized. Therefore, how to intelligently deconstruct the existing complete course video and automatically extract knowledge point-level micro course resources to support flexible resource reorganization and personalized supply becomes a technical problem to be solved in the current digital education resource construction field. Disclosure of Invention In view of the above, the application provides a micro-class extraction method, a related method and a related device, which are used for intelligently deconstructing the existing complete course video and automatically extracting knowledge point-level micro-classes so as to support flexible resource recombination and personalized supply, and the technical scheme is as follows: A micro class extraction method, comprising: acquiring a target course video of a target subject; analyzing the content of the target course video to generate corresponding target text data with a time stamp; Carrying out knowledge point slicing on the target text data based on a large slicing model to obtain knowledge point slicing results, wherein the knowledge point slicing results at least comprise start and stop time of knowledge point fragments, the large slicing model adopts training text data marked with real knowledge point slicing results and is supplemented with thinking process data for obtaining the real knowledge point slicing results, and a universal large model base is trained to obtain the training text data which is text data with time stamps and corresponds to course video samples of the target subjects; and extracting knowledge point video clips from the target course video according to the start-stop time of the knowledge point clips to obtain knowledge point video micro-courses. In one possible implementation manner, the parsing the content of the target course video to generate corresponding target text data with a timestamp includes: performing audio and video separation on the target course video to obtain audio data and silent video data; and processing the audio data into standard, correct and structured text data with time stamps and speaker labels by combining the video content of the silent video data to obtain target text data corresponding to the target course video. In one possible implementation, the processing the audio data into canonical, correct, time-stamped and speaker-tagged structured text data in combination with the video content of the silent video data includes: performing voice recognition on the audio data to obtain a recognition text with a time stamp, and distinguishing the teacher utterance from the student utterance by the recognition text to obtain a text with the time stamp and a speaker tag; extracting a key video frame from the silent video data, and extracting text information from the key video frame; Based on a large language model, combining text information extracted from the key video frames, correcting semantic, grammar and key information of the text with the timestamp and the speaker tag, and performing long sentence segmentation processing, formula standardization processing and semantic-based paragraph division processing on the corrected text to obtain a structured text with the standard, correct, timestamp and the speaker tag. In one possible implementation