Video resource knowledge point labeling method and system, computer equipment and medium

Abstract

The invention relates to the technical field of video processing, and discloses a knowledge point labeling method and system for video resources, computer equipment and a medium. The method comprises the following steps: determining each clause corresponding to an audio file of a video resource to be labeled in a recording and broadcasting system; determining first timestamp information corresponding to each clause, and sequentially labeling to obtain a corresponding line number; determining a first mapping data table according to the corresponding relationship between the first timestamp information and the line number; determining a teaching outline according to each clause and the line number corresponding to each clause; determining second timestamp information corresponding to each group of prompt words under the teaching outline according to the first mapping data table; constructing index information according to the second timestamp information and the knowledge point information corresponding to each group of cue words; and labeling the to-be-labeled video resource according to the index information. According to the method, the index information based on the knowledge points and the timestamps is constructed, so that the user can accurately position the content of the knowledge points in the video, and the search efficiency and accuracy of the video are improved.

Description

Technical Field

[0001] The present invention relates to the field of video processing technology, and in particular to a method, system, computer equipment and medium for labeling knowledge points of video resources. Background Art

[0002] The regular recording and broadcasting system is widely used in schools. It can record data such as the class location, class time, lecturer, course name, etc.

[0003] The recording and broadcasting system records some basic data. Since video resources are different from text-based teaching resources, searching for their content consumes huge computing power and a lot of the inquirer's time. Simply searching based on some basic data is difficult to efficiently and accurately find the required course recordings from the massive video library.

[0004] Therefore, there is an urgent need for a video resource annotation method that can improve the search efficiency and accuracy of target video resources. Summary of the invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the problem in the related art that it is difficult to efficiently and accurately find the required course recordings from a massive video library by searching based on partial basic data.

[0006] In order to solve the above technical problems, the present invention provides a knowledge point labeling method for video resources, which is applied to a recording and broadcasting system. The knowledge point labeling method for video resources includes:

[0007] Determine each sentence corresponding to the audio file of the video resource to be annotated in the recording and broadcasting system;

[0008] Determine the first timestamp information corresponding to each sentence, and label them in sequence according to the sentence order to obtain the line number corresponding to each sentence; the first timestamp information corresponding to each sentence includes: a first start timestamp and a first end timestamp corresponding to each sentence;

[0009] Determine a first mapping data table according to the correspondence between the first timestamp information corresponding to each clause and the row number;

[0010] Determine a teaching outline according to each sentence and the line number corresponding to each sentence; the teaching outline includes: at least one group of prompt words, each group of prompt words includes a line number range and knowledge point information stored in a corresponding relationship, and the knowledge point information is determined by analyzing the knowledge points of each sentence;

[0011] Determine the second timestamp information corresponding to each group of prompt words under the teaching outline according to the first mapping data table; the second timestamp information corresponding to each group of prompt words includes: a second start timestamp and a second end timestamp corresponding to each group of prompt words; the second start timestamp is the first start timestamp corresponding to the smallest row number in the row number range, and the second end timestamp is the first end timestamp corresponding to the largest row number in the row number range;

[0012] Constructing index information according to the second timestamp information corresponding to each group of prompt words and the knowledge point information in each group of prompt words; a single piece of index information includes the second timestamp information corresponding to a single group of prompt words and the knowledge point information in the group of prompt words;

[0013] The video resource to be marked is marked according to the index information.

[0014] In an optional implementation manner, the determining each sentence corresponding to the audio file of the video resource to be annotated in the recording and broadcasting system includes:

[0015] An audio file of a video resource to be annotated in a recording and broadcasting system is determined, text information corresponding to the audio file is determined, and the text information is split according to a preset sentence separator to obtain individual sentences.

[0016] In an optional implementation, determining the teaching outline according to each sentence and the line number corresponding to each sentence includes:

[0017] Pre-build a subject knowledge point repository based on the knowledge point information in each subject textbook; the subject knowledge point repository includes the knowledge point information in the subject textbook corresponding to the video resource to be annotated;

[0018] Analyze each sentence to obtain initial prompt words; a single group of initial prompt words includes an initial line number range and initial knowledge point information stored in a corresponding relationship;

[0019] The initial prompt words are written into a preset prompt word assembly template to obtain an initial outline, and the initial outline is adjusted according to the subject knowledge point repository to obtain the teaching outline.

[0020] In an optional implementation, determining the teaching outline according to each sentence and the line number corresponding to each sentence includes:

[0021] The teaching outline is determined by a pre-built large model based on each sentence and the line number corresponding to each sentence.

[0022] In an optional implementation, after determining the teaching outline according to each sentence and the line number corresponding to each sentence, the method further includes:

[0023] Determine whether there are invalid results in each group of prompt words under the teaching outline; the invalid results indicate that the preset prompt word assembly template is not satisfied and includes a line number range;

[0024] When invalid results exist in each group of prompt words under the teaching outline, the invalid results are adjusted until the invalid results meet the preset prompt word assembly template.

[0025] In an optional implementation, the line numbers in the line number range corresponding to each group of prompt words under the teaching outline are consecutive numbers.

[0026] In an optional embodiment, the method further includes:

[0027] A saving interface is set in the recording and broadcasting system; the saving interface is used to save all index information of the video resource to be annotated.

[0028] In a second aspect, the present invention provides a knowledge point annotation system for video resources, which is applied to a recording and broadcasting system. The knowledge point annotation system for video resources includes:

[0029] The first processing module is used to determine each sentence corresponding to the audio file of the video resource to be annotated in the recording and broadcasting system;

[0030] The second processing module is used to determine the first timestamp information corresponding to each sentence, and label the sentences in sequence to obtain the line number corresponding to each sentence; the first timestamp information corresponding to each sentence includes: the first start timestamp and the first end timestamp corresponding to each sentence; determine the first mapping data table according to the correspondence between the first timestamp information corresponding to each sentence and the line number;

[0031] The third processing module is used to determine the teaching outline according to each sentence and the line number corresponding to each sentence; the teaching outline includes: at least one group of prompt words, each group of prompt words includes a line number range and knowledge point information stored in a corresponding relationship, and the knowledge point information is determined by analyzing the knowledge points of each sentence;

[0032] A fourth processing module is used to determine the second timestamp information corresponding to each group of prompt words under the teaching outline according to the first mapping data table; the second timestamp information corresponding to each group of prompt words includes: a second start timestamp and a second end timestamp corresponding to each group of prompt words; the second start timestamp is the first start timestamp corresponding to the smallest row number in the row number range, and the second end timestamp is the first end timestamp corresponding to the largest row number in the row number range;

[0033] A fifth processing module, configured to construct index information according to the second timestamp information corresponding to each group of prompt words and the knowledge point information in each group of prompt words; a single piece of index information includes the second timestamp information corresponding to a single group of prompt words and the knowledge point information in the group of prompt words;

[0034] The sixth processing module is used to mark the video resources to be marked according to the index information.

[0035] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the knowledge point labeling method for video resources of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0036] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a single computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the knowledge point labeling method for video resources of the first aspect or any corresponding embodiment thereof.

[0037] In a fifth aspect, the present invention provides a computer program product, including computer instructions, which are used to enable a computer to execute the knowledge point labeling method for video resources of the first aspect or any corresponding embodiment thereof.

[0038] The technical solution provided by the present invention has the following technical effects:

[0039] Precise positioning and navigation: Fine-grained positioning based on timestamps: By determining the first timestamp information for each sentence and constructing the first mapping data table, the video content can be accurately time-positioned at the sentence level. This allows users to jump directly to the start time of the first sentence of the corresponding knowledge point when querying a knowledge point, avoiding the inconvenience of traditional video positioning that can only rely on rough timeline dragging, greatly improving the accuracy and efficiency of positioning. For example, when learning a specific "steps for solving quadratic equations" knowledge point in a math course video, you can quickly locate the video clip where the sentence explaining the knowledge point is located, saving a lot of search and positioning time.

[0040] Navigation based on the syllabus: The syllabus is presented in the form of line number range and knowledge point information, providing users with a clear navigation of the video content structure. Users can first browse the syllabus to understand the overall picture of the knowledge points covered in the video, and then select the knowledge points of interest and their corresponding video clips for study according to their own needs. This method is similar to the directory navigation of a book, which helps users quickly grasp the core content of the video and improve the retrieval efficiency and accuracy of the video corresponding to the knowledge point.

[0041] Knowledge point extraction and integration: Analyze and determine the knowledge point information from each sentence and build it into the teaching outline, realizing automatic knowledge extraction and organization of video content, and improving the efficiency of knowledge point information annotation. Based on the constructed index information, the video resources can be annotated to decompose the long video into multiple video clips according to the knowledge points. Each video clip explains different knowledge points, which makes it convenient for users to find the required course recordings from the massive video library efficiently and accurately according to their own needs.

[0042] The technical solution of the present invention integrates multiple technical means such as audio sentence processing, timestamp marking, natural language processing (for knowledge point analysis), data mapping and index construction, and realizes the organic coordination of various technical links. The practical problem of video resource knowledge point annotation is solved through the innovative model of multi-technology integration. The technical solution of the present invention is applied to the recording and broadcasting system, and can be easily integrated with the video recording, storage, playback and other functional modules related to the recording and broadcasting system. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related technologies, the drawings required for use in the specific embodiments or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 is a flow chart of a method for labeling knowledge points of video resources according to an embodiment of the present invention;

[0045] Figure 2 is a schematic diagram of a specific process of a method for labeling knowledge points of video resources according to an embodiment of the present invention;

[0046] Figure 3 is a schematic diagram of the structure of a knowledge point annotation system for video resources according to an embodiment of the present invention;

[0047] Figure 4 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0049] The normalized recording and broadcasting system is widely used in schools. It can record basic data such as the location of the class, class time, lecturer, course name, etc. The recording and broadcasting system can record some basic data, but the lecture content information needs to be supplemented by the lecturer to the recording and broadcasting system after the recording is completed. Most recorded videos rely on teachers to supplement the lecture content data later. Teachers need to supplement it after class, and they are prone to forget to supplement it, and there will be omissions or marking errors. As a result, this data is often lacking. However, course content data is the key to querying recorded course content. Since video resources are different from text-based teaching resources, searching for their content consumes huge computing power and a lot of time for the queryer. Therefore, it is difficult to efficiently and accurately find the required course recordings from the massive video library. There is an urgent need for the recording and broadcasting system to automatically extract video teaching knowledge points to improve this situation.

[0050] To this end, the embodiments of the present invention provide a method, system, computer device and medium for labeling knowledge points of video resources to solve the problem in related technologies that it is difficult to efficiently and accurately find the required course recordings from a massive video library by searching based on partial basic data.

[0051] According to an embodiment of the present invention, an embodiment of a method for labeling knowledge points of video resources is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer device such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0052] Figure 1 It is a flowchart of a method for labeling knowledge points of video resources according to an embodiment of the present invention.

[0053] like Figure 1 As shown, an embodiment of the present invention provides a method for labeling knowledge points of video resources, which is applied to a recording and broadcasting system. Based on the technical solution of the present invention, the recording and broadcasting system can realize automatic labeling of knowledge points of video resources.

[0054] The knowledge point annotation method of the video resource includes:

[0055] S101: Determine each sentence corresponding to the audio file of the video resource to be annotated in the recording and broadcasting system.

[0056] In this embodiment, S101 determines each sentence corresponding to the audio file of the video resource to be annotated in the recording and broadcasting system, specifically including:

[0057] An audio file of a video resource to be annotated in a recording and broadcasting system is determined, text information corresponding to the audio file is determined, and the text information is split according to a preset sentence separator to obtain individual sentences.

[0058] As a specific example, the audio file of the video resource to be annotated in the recording and broadcasting system can be extracted by an audio extraction tool. As an example, the audio extraction tool can be an FFMpeg tool. The text information corresponding to the audio file can be determined by speech recognition technology (Automatic Speech Recognition, ASR). The text information can be split according to preset sentence separators, such as semicolons, periods, question marks and other ending symbols, to obtain individual sentences. The splitting process can be to scan from the starting position of the text information, and separate the two sentences when the symbol identical to the preset sentence separator is scanned, and the text content from the beginning to the preset sentence separator is determined as a sentence, and the next word of the current preset sentence separator is rescanned as the beginning, and so on, and multiple sentences can be separated sentence by sentence.

[0059] As another specific example, the following is a specific implementation method of determining an audio file of a video resource to be annotated in a recording and broadcasting system, converting the audio file into text information, and then splitting the text information into individual sentences according to preset sentence separators:

[0060] Get audio files: In the recording and broadcasting system, there is usually a corresponding storage module to manage video resources and the corresponding audio files. First, locate the corresponding audio file through the file index of the recording and broadcasting system or the video resource identifier specified by the user (such as video file name, unique number, etc.). For example, if the video resources are stored according to the course name and date, the audio file to be processed can be determined based on the association method such as "Mathematics Course_20241215.mp4" corresponding to the audio file "Mathematics Course_20241215.wav".

[0061] Audio to text (speech recognition): Use speech recognition technology to convert audio files into text information. Common implementation methods are as follows:

[0062] Choose a speech recognition tool or service: You can use an open source speech recognition library, such as CMUSphinx, which supports multiple languages. After installing the corresponding language model and configuring the recognition parameters (such as sampling rate, audio format, etc.), you can input the audio file and output the corresponding text information. You can also use the speech recognition API provided by the cloud service platform to convert the audio file into text information.

[0063] Parameter setting and optimization during speech recognition: Audio preprocessing: Preprocess the audio files according to the requirements of the selected speech recognition tool, such as unifying the audio sampling rate (common ones are 16000Hz, etc.), the number of channels (mono or dual-channel conversion, etc.), removing noise interference in the audio (the audio processing library can be used for noise reduction), ensuring that the audio quality meets the optimal conditions for speech recognition and improving recognition accuracy. Language model selection and adaptation: If the speech recognition tool supports multiple language models, select a model that matches the language of the audio content. For example, for Mandarin Chinese teaching videos, select the corresponding Chinese language model. And for specific fields (such as mathematics, physics, etc.), some speech recognition tools also allow the use of domain adaptation technology to further optimize the language model so that the recognized text is more in line with professional content expression habits.

[0064] Text information splitting (based on preset sentence separators): After obtaining the text information corresponding to the audio file, split it according to the preset sentence separators to obtain each sentence. The specific operations are as follows:

[0065] Determine the preset sentence separators: Commonly used preset sentence separators generally include punctuation marks, such as the period ("."), question mark ("."), and exclamation mark ("."), which often mark the end of a sentence in most languages ​​and can be used as natural sentence boundaries. In addition, for some specific fields or texts with special formatting requirements, other symbols can be added as sentence separators. For example, in code-related texts, the semicolon (".") is also often used as a statement end marker. In mathematical formula expression texts, customize some special symbols to assist sentence separation according to the logical structure of the formula (such as each equation and inequality is a separate sentence, etc.).

[0066] S102: Determine the first timestamp information corresponding to each sentence, and label them in sequence according to the sentence order to obtain the line number corresponding to each sentence, and determine the first mapping data table according to the correspondence between the first timestamp information corresponding to each sentence and the line number.

[0067] In this embodiment, the first timestamp information corresponding to each sentence includes: a first start timestamp and a first end timestamp corresponding to each sentence.

[0068] In this embodiment, in the process of converting the audio file into corresponding text information, the first timestamp information corresponding to each sentence can be determined according to the start time and end time of each sentence in the audio file.

[0069] As an example, when converting an audio file into corresponding text information with the help of speech recognition technology, the speech recognition tool generally provides a mechanism to obtain the time information corresponding to the speech content. For example, many advanced speech recognition algorithms process audio based on the time series of the audio when analyzing the audio, and record the start and end time of each speech segment (corresponding to the subsequent split sentences) in the original audio.

[0070] In this embodiment, the line numbers corresponding to the sentences can be obtained by sequentially labeling from a preset value, such as 1, according to the sentence order or time order. After obtaining the first timestamp information corresponding to each sentence, a number is assigned to each sentence starting from 1 according to the order in which the sentences appear in the audio file (or text information). This number is the line number.

[0071] In this embodiment, the purpose of the first mapping data table is to clearly present the correspondence between the first timestamp information and the row number corresponding to each clause. A variety of data structures can be used to construct the first mapping data table. As an example, a data dictionary can be used to construct the first mapping data table. In the data dictionary, the key is the row number, and the corresponding value is a sub-dictionary containing the first start timestamp and the first end timestamp of the clause, which can easily query the first timestamp information corresponding to any row number, or find the corresponding row number according to the first timestamp range, etc., which provides basic data support for subsequent operations such as knowledge point annotation based on timestamps and row numbers. In this embodiment, a suitable data structure (such as a database table structure, etc.) can be flexibly selected according to actual needs to construct and store this first mapping data table, but the core idea is to reflect the clear correspondence between the first timestamp information and the row number.

[0072] In this embodiment, the first mapping data table may also include text content of each sentence.

[0073] As a specific example, L1: (first start timestamp: 0s, first end timestamp: 20s), L2: (first start timestamp: 21s, first end timestamp: 30s), L2: (first start timestamp: 32s, first end timestamp: 50s).

[0074] As a specific example, when the first mapping data table may further include text content of each sentence, the first mapping data table may be:

[0075] 1: Hello students, today we will talk about quadratic equations (start: 1, end: 20).

[0076] 2: There are two parameters in the quadratic equation, x and y (start: 21, end: 30).

[0077] 3: Let's look at an example (start: 31, end: 50).

[0078] S103: Determine the teaching outline according to each sentence and the line number corresponding to each sentence.

[0079] In this embodiment, the teaching outline includes: at least one group of prompt words, a single group of prompt words includes a line number range and knowledge point information stored in a corresponding relationship, and the knowledge point information is determined by analyzing the knowledge points of each sentence. The knowledge point information may include the name of the knowledge point, the outline title, the outline summary, etc.

[0080] In this embodiment, the number of prompt words included in the teaching outline can be limited, for example, 10 groups, and the specific value can be set and modified according to actual needs, and is not limited here. A single group of prompt words usually includes multiple sentences, and the line numbers in the line number range corresponding to each group of prompt words under the teaching outline are consecutive numbers.

[0081] In this embodiment, a subject knowledge point repository can be pre-constructed based on the knowledge point information in each subject textbook. The subject knowledge point repository includes the knowledge point information in the subject textbook corresponding to the video resource to be annotated. Each sentence is matched to determine the knowledge point information in each sentence that matches the subject knowledge point repository as the knowledge point information corresponding to the sentence, and the sentences with the same knowledge point information and continuous line numbers and the corresponding line number range are determined as a group of prompt words. For example, if the keyword "quadratic equation" appears in a sentence, it can be preliminarily determined that the sentence involves the knowledge point of "quadratic equation".

[0082] In the present embodiment, for each subject, the authoritative teaching materials of the subject are collected. For example, for the subject of mathematics, mathematics teaching materials from primary school to different stages of university are collected. Professionals (such as subject teachers or education experts) are arranged to read the teaching materials carefully, and comb out the knowledge point information chapter by chapter and section by section. For the contents such as charts, examples in the teaching materials, the knowledge points contained therein are also extracted. A mechanism for regularly updating the subject knowledge point repository can also be established. When the subject teaching materials are updated, the curriculum standards are revised, or new academic research results appear, the repository is updated in time. A quality review process can be established for the newly added or modified knowledge points. An audit team can be formed by subject experts to audit the accuracy, completeness and applicability of the knowledge points. After the audit is passed, the knowledge points can be officially updated to the repository to ensure that the knowledge points in the repository always maintain high quality and authority.

[0083] In this embodiment, corresponding subject knowledge point repositories can be pre-built for different subjects (such as mathematics, physics, etc.), covering important terms such as common concepts, principles, methods, etc. of the subject.

[0084] S104: Determine the second timestamp information corresponding to each group of prompt words in the teaching outline according to the first mapping data table.

[0085] In this embodiment, the second timestamp information corresponding to each group of prompt words includes: a second start timestamp and a second end timestamp corresponding to each group of prompt words. The second start timestamp is the first start timestamp corresponding to the smallest line number in the line number range, and the second end timestamp is the first end timestamp corresponding to the largest line number in the line number range.

[0086] For example, the second start timestamp of the row number range [1, 2, 3] is the first start timestamp corresponding to the smallest row number 1, and the second end timestamp is the second end timestamp corresponding to the largest row number 3.

[0087] S105: constructing index information according to the second timestamp information corresponding to each group of prompt words and the knowledge point information in each group of prompt words.

[0088] In this embodiment, a single piece of index information includes the second timestamp information corresponding to a single group of prompt words and the knowledge point information in the group of prompt words. Usually, a group of prompt words can construct a piece of index information.

[0089] In this embodiment, index information can be constructed according to the second timestamp information corresponding to each group of prompt words and the knowledge point name in the knowledge point information. As an example, quadratic equation: [1, 2], basic concept of set: [3, 4, 5, 6]. According to the first mapping data table, the corresponding index information can be determined as: quadratic equation: [0, 30], basic concept of set: [30, 60].

[0090] S106: Label the video resource to be labeled according to the index information.

[0091] In this embodiment, a saving interface may be set in the recording and broadcasting system. The saving interface is used to save all index information of the video resources to be annotated.

[0092] In this embodiment, the specific implementation method of marking the video resources to be marked according to the index information is as follows:

[0093] Select annotation tools and techniques: Video editing software: Professional video editing software such as Adobe Premiere Pro and Final Cut Pro provide a wealth of annotation functions. Annotation can be achieved by writing scripts or using the software's own marking tools. For example, in Adobe Premiere Pro, using the marking function on its timeline, add a mark at the corresponding time point according to the second timestamp information in the index information, and use the knowledge point information as the annotation content of the mark. In this way, when the video is playing, when the play pointer reaches the marked position, the corresponding knowledge point information can be displayed, which is convenient for users to understand the content of the video here.

[0094] Annotation function based on the recording and broadcasting system: If the video resources are stored on a specific video platform (such as an online education platform, an internal corporate video library platform, etc.), the recording and broadcasting system itself provides a video annotation interface or functional module. Developers can pass the knowledge points and the second timestamp information in the index information according to the Application Programming Interface (API) provided by the platform to realize automatic annotation of the video on the recording and broadcasting system. For example, in the video management background of an online education platform, by calling the annotation interface and passing in parameters such as the knowledge point name, start timestamp, and end timestamp, the recording and broadcasting system will display the knowledge point information at the corresponding time position on the video playback page, making it easier for students to quickly locate and understand the key content when watching video courses.

[0095] In addition to directly displaying index information during video playback, you can also mark it on the video progress bar. For example, according to the timestamp in the index information, add a mark point or a line segment of different colors to the corresponding position of the progress bar. When the user hovers the mouse over the marked position of the progress bar, the detailed information of the knowledge point pops up.

[0096] In this embodiment, an interactive function may be provided in the recording and broadcasting system.

[0097] Adding interactive functions: In order to enhance the interactivity between users and index information, some operation functions can be added. For example, users can jump to the video of the relevant knowledge point by clicking on the index text box or the mark on the progress bar. When users query videos in the recording and broadcasting system, the recording and broadcasting system provides subject knowledge point selection conditions. After the user selects the knowledge point, the recording and broadcasting system uses the constructed index information to query the video and provide it to the user.

[0098] When the user receives a video file, he / she also receives all the index information, knowledge point names, and start and end times of each knowledge point corresponding to the video. The knowledge point list is displayed on the user playback interface. When the user clicks on a knowledge point name, the player will automatically jump to the corresponding time point for playback based on the second start timestamp of the knowledge point.

[0099] In an optional implementation, S103 determines the teaching outline according to each sentence and the line number corresponding to each sentence, specifically including:

[0100] S1031: Pre-build a subject knowledge point repository based on the knowledge point information in the textbooks of each subject.

[0101] In this embodiment, the subject knowledge point storage library includes knowledge point information in the subject textbook corresponding to the video resource to be annotated.

[0102] In this embodiment, in order to ensure the accuracy of the knowledge points output by the large model, the recording and broadcasting system presets the knowledge point information of each subject textbook. Through the basic teaching information of the recording and broadcasting system, the subject knowledge points are queried and assembled into the large model prompt words.

[0103] S1032: Analyze each sentence to obtain an initial prompt word.

[0104] In this embodiment, a single group of initial prompt words includes an initial line number range and initial knowledge point information stored in a corresponding relationship.

[0105] In this embodiment, as an example, a natural language processing (NLP) analysis is performed on each sentence. A part-of-speech tagging tool (such as the part-of-speech tagging function in the Natural Language Toolkit (NLTK) library) can be used to determine the key words such as nouns and verbs in each sentence. For example, for the sentence "the root-finding formula of a quadratic equation is very important", key information such as "quadratic equation" (noun phrase) and "root-finding formula" (noun phrase) can be identified through part-of-speech tagging. The professional terms or specific entities therein are further determined by using named entity recognition (NER) technology. In the above example, it is accurately identified that "quadratic equation" is a specific knowledge point entity in the subject of mathematics. Based on these analysis results, the knowledge point information that may be involved in each sentence is preliminarily determined, and an initial line number range is assigned. For example, if the clauses involved in the knowledge point information are the 5th and 6th clauses, the initial line number range can be temporarily set to [5, 6]. Combine this information to form an initial prompt word, for example: {[5, 6]: "The formula for finding the roots of a quadratic equation"}.

[0106] S1033: Writing the initial prompt words into a preset prompt word assembly template to obtain an initial outline, and adjusting the initial outline according to the subject knowledge point repository to obtain a teaching outline.

[0107] In this embodiment, each sentence, the line number corresponding to each sentence, the preset prompt word assembly template and the subject knowledge point storage library can be used as the input data of the large model, and the teaching outline is the output result of the large model. The large model is used to execute the technical solution of determining the teaching outline according to each sentence and the line number corresponding to each sentence.

[0108] In this embodiment, a large model commonly used in natural language processing in the field can be used as the initial model, and targeted training can be performed according to application requirements to obtain the large model in the technical solution of the present invention. The present invention can use the conventional model training method in the field to train and tune the large model, and does not limit the specific process of model training and tuning.

[0109] As an example, here are the high-level steps to build and train a large model to achieve the above functionality:

[0110] Building a large model:

[0111] Determine the model architecture: You can choose an infrastructure such as the Transformer architecture, which has been widely proven to perform well in natural language processing. The architecture model adopted by the common BERT and GPT series models has the advantages of parallel computing capabilities and the ability to capture semantic dependencies in long sequences of text. It is suitable for processing text information in the recording system to generate teaching outlines.

[0112] Consider the model's hyperparameter settings such as the number of layers, hidden layer dimensions, number of heads (involved in the multi-head attention mechanism in the Transformer architecture) according to specific needs. For example, for more complex multidisciplinary and long-length lecture content processing scenarios, it may be necessary to appropriately increase the number of layers and hidden layer dimensions to improve the model's representation capabilities.

[0113] Data preparation and preprocessing:

[0114] Collecting corpus: In addition to organizing and collecting the knowledge point information of each subject textbook, it is also necessary to collect a large number of relevant correct teaching outline example texts with different styles and expressions, as well as data similar to actual application scenarios containing various sentence forms, corresponding line number markings, etc. These data sources can be lesson plans written by excellent teachers in the past, audited standard teaching outline documents, etc.

[0115] Text cleaning: Normalize the collected text, such as removing extra spaces, punctuation errors, unifying uppercase and lowercase formats, etc. At the same time, eliminate data of poor quality (such as incomplete content, confusing logic, etc.) to ensure the data quality of the input model.

[0116] Word segmentation and tokenization: Use the selected natural language processing tools (such as the commonly used NLTK, Jieba and other word segmentation libraries, and select appropriate ones for different languages) to segment the text into words or sub-word units, and perform corresponding token conversion to facilitate subsequent recognition and processing by the model, such as mapping words in the text into digital indexes and other forms.

[0117] Data partitioning: Divide the processed data into training set, validation set and test set according to a certain ratio. The common partition ratio can be 8:1:1 or 7:2:1, etc. The validation set is used to adjust the model hyperparameters and evaluate the performance of the model on unseen data during the training process. The test set is used for the final objective evaluation of the model's generalization ability.

[0118] Training a large model:

[0119] Choose the appropriate training algorithm and optimizer:

[0120] Training algorithm: Common supervised learning algorithms can be used in this scenario because there are clear inputs (sentences, line numbers, knowledge point libraries, etc.) and corresponding expected outputs (lecture outlines). For example, algorithms based on gradient descent can be used, such as Stochastic Gradient Descent (SGD) and its variants Adagrad, Adadelta, Adam, etc. The Adam optimizer is widely used in many natural language processing tasks due to its advantages such as adaptive learning rate adjustment. It can be considered to accelerate the model training convergence process.

[0121] Define the loss function: Select an appropriate loss function based on the task objectives to measure the difference between the lecture outline predicted by the model and the actual lecture outline. For example, the cross entropy loss function can be used (commonly used in multi-classification text output scenarios when processing classification or generation tasks). By minimizing the loss function value, the model parameters are continuously optimized so that they are updated in the direction of outputting a more accurate lecture outline.

[0122] Start the training process: input the data in the training set into the constructed model in batches. After each batch of data is input, the model performs forward propagation calculation based on the current parameters to predict the output, and then compares the predicted result with the true label (that is, the actual syllabus) to calculate the loss value. Then, through the back propagation algorithm, the model parameters are updated according to the calculated gradient information. This process is repeated continuously. After multiple rounds of training (epochs), observe the changes in the performance indicators of the model on the validation set (such as accuracy, recall rate, F1 value, etc., which are used to evaluate the accuracy of syllabus generation). Adjust the hyperparameters (such as learning rate, batch size, etc.) in a timely manner to prevent overfitting (which can be achieved through technical means such as early stopping and regularization) until the performance of the model on the validation set reaches a better and stable state.

[0123] Model evaluation and tuning: Use the test set to conduct a final evaluation of the trained model to obtain its performance on real unseen data. If the results are not ideal, analyze the possible problems, such as insufficient data volume, insufficient model complexity, etc., further supplement the data, adjust the model architecture or hyperparameters, and then retrain and evaluate, and continuously iterate and optimize until the model can accurately and reliably generate a teaching outline that meets the requirements based on the input sentence, line number and other information. It is worth noting that building and training high-quality large models is a complex and resource-intensive process that requires the support of powerful computing resources (such as GPU clusters, etc.), and the performance of the model must be continuously improved based on actual application feedback.

[0124] In this embodiment, the knowledge point information in each group of prompt words under the teaching outline can be jointly determined based on the knowledge point information in each sentence and the knowledge point information in the subject knowledge point storage library, so that the obtained teaching outline is more accurate.

[0125] In this embodiment, the teaching outline is obtained by adjusting the initial outline according to the subject knowledge point repository. Specifically, the knowledge point information and the line number range can be adjusted. For example, the specific description of the knowledge point information is adjusted, prompt words are integrated, and the line number range and knowledge point information are integrated. When multiple line number ranges correspond to one knowledge point information, the multiple line number ranges are integrated into one line number range. When the line number range is large, the knowledge point information is refined through further analysis and matching. Multiple knowledge point information is determined through hierarchical processing, and the line number range is correspondingly split. The knowledge point information can be further refined, which is convenient for building more index information in the later stage, and the video resources can be segmented in more detail.

[0126] In this embodiment, the initial prompt words obtained by the previous analysis are filled in according to the template requirements. For example, the subject name (assuming it is mathematics) is filled in the first %s position, and then the knowledge point information and line number range in the initial prompt words are filled in the corresponding positions in the template in sequence according to the format requirements, and a similar result is obtained: "You are a mathematics teacher, strictly follow the order of the article to generate an outline for the following content. Requirements: Summarize the knowledge points of the root-finding formula of the quadratic equation described in the text. The line numbers of each part must be continuous when summarizing the outline. The generated outline should not exceed 10 parts. Generate according to the teaching content, and ignore the rest of the content that is not related to the teaching. Each part of the summary outline needs to be in the following format: <<[Line H: 5-H: 6] TITLE: Root-finding formula of quadratic equation-INFO: Explain the application of the root-finding formula of quadratic equation-KNOW: Root-finding formula of quadratic equation. >> It should be noted that each part of the outline must start with << and end with >>". In this way, the initial outline is obtained, and the initial outline may contain multiple parts composed of prompt words generated according to different sentences.

[0127] As an example, the initial outline is adjusted according to the subject knowledge point repository to obtain the teaching outline:

[0128] Knowledge point matching and integration: traverse each prompt word part in the initial outline. For the initial knowledge point information in each part, perform an exact match in the subject knowledge point repository. For example, there is a prompt word part about "the formula for finding the roots of a quadratic equation" in the initial outline. Look for the standard knowledge point expression that fully matches or is closest to it in the subject knowledge point repository (such as "quadratic equation-root formula" in the repository). If a matching knowledge point is found, check the knowledge system classification to which it belongs, other related knowledge points and other information, and optimize the prompt word part in the initial outline. For example, adjust the outline summary of the INFO part to make it more consistent with the standard description and teaching requirements of the knowledge point in the subject knowledge point repository. It may be possible to adjust "explain the application of the formula for finding the roots of a quadratic equation" to "in-depth explanation of the principles and application scenarios of the formula for finding the roots of a quadratic equation". If the knowledge point in the initial outline does not find an exact match in the subject knowledge point repository, further semantic analysis and manual review (if necessary) are performed. For example, the initial knowledge point information is "quick solution to special quadratic equations", which is not directly matched in the repository. Through manual analysis, it is determined that it is a special case of "solution to quadratic equations". It is integrated into the corresponding standard knowledge point framework, and the prompt word part in the initial outline is updated. Its line number range and other information are reasonably merged or adjusted with the prompt word part of the relevant standard knowledge points to ensure the logic and coherence of the entire outline.

[0129] Outline structure optimization: Check whether the line number ranges of each prompt word part in the initial outline are continuous and reasonable. If there are unreasonable jumps or overlaps in the line number range, adjust them according to the sentence order and the logical relationship of the knowledge points. For example, the line number ranges of the two relevant prompt word parts are [3, 3] and [5, 5] respectively, but the knowledge point of the fourth sentence in the middle is closely related to these two parts. Adjust the line number range to [3, 4, 5] and integrate the corresponding knowledge point information and outline description content.

[0130] At the same time, according to the requirement that the generated outline should not exceed 10 parts, the initial outline is streamlined and merged. If there are some overly trivial or similar prompt word parts, they are merged into a more comprehensive part to ensure that the final teaching outline not only covers the key knowledge points in the video content, but also has the characteristics of being concise, clear and well-structured, which is convenient for teachers to teach and students to learn.

[0131] In this embodiment, a preset prompt word assembly template can be used to put text information and line numbers into prompt words to obtain initial prompt words. Specifically, the teaching outline can be obtained by assembling through the Prompt assembly engine.

[0132] As an example, you can use a preset prompt word assembly template, put the text information and line number into the prompt word to get the initial prompt word. The recording and broadcasting system presets the knowledge point information of each subject textbook, and queries the subject knowledge points through the basic teaching information of the recording and broadcasting system and assembles them into the large model prompt word. According to the knowledge point information in the text information and the prompt word information merged from the subject knowledge point repository, call the large model, assemble it using the string template engine, pass in four parameters: teaching subject, teaching chapter, subject knowledge point, and classroom content, and give the result template in the prompt word of the large model, indicating that the large model must return the line number range in the result. The assembled prompt word information and other parameters are passed into the large model, and the large model generates a teaching outline for the text information according to the prompt word requirements, and returns the result in the specified format.

[0133] An example of the result (lecture outline) returned by the large model is as follows:

[0134] <<[Line H:1-H:2]title: Quadratic equations - INFO: The course talks about quadratic equations - KNOW: How to solve quadratic equations>>;

[0135] <<[Lines H:3-H:6]title: Sets in equations -INFO: The course explains the basic concepts of sets -KNOW: The concept of sets>>.

[0136] In an optional implementation, S103 determines the teaching outline according to each sentence and the line number corresponding to each sentence, specifically including:

[0137] The teaching outline is determined by a pre-built large model based on each sentence and the line number corresponding to each sentence.

[0138] In this embodiment, a large model is used to automatically annotate knowledge point information to improve the speed of searching for video resources in a normalized recording and broadcasting system.

[0139] In this embodiment, the prompt word assembly template can be:

[0140] <<[Line H:mH:n] TITLE: Outline title within 30 words - INFO: Outline summary, within 50 to 150 words - KNOW: Knowledge points included in the outline. >>.

[0141] In this example, you are a %s teacher. Strictly follow the order of the article to generate an outline for the following content. Requirements: Summarize the %s knowledge points described in the text (text information). The line numbers of each part in the outline must be continuous. The generated outline should not exceed 10 parts. Generate according to the teaching content, and ignore the rest of the content that is not related to teaching. Each part of the outline needs to follow the following format:

[0142] <<[Line H:mH:n] TITLE: Outline title within 30 words - INFO: Outline summary, within 50 to 150 words - KNOW: Knowledge points included in the outline. >>.

[0143] It should be noted that each part of the outline must start with << and end with >>.

[0144] Note: The knowledge point information contained in the outline needs to be selected from the following knowledge point information that best matches: The knowledge point information is as follows: %s. The text information contained in the outline is as follows: %s.

[0145] In an optional implementation, after determining the teaching outline according to each sentence and the line number corresponding to each sentence in S103, the knowledge point marking method of the video resource further includes:

[0146] Determine whether there are invalid results in each group of prompt words under the course outline.

[0147] In this embodiment, the invalid result indicates that the preset prompt word assembly template is not satisfied and includes a line number range.

[0148] When there are invalid results in each group of prompt words under the teaching outline, the invalid results are adjusted until the invalid results meet the preset prompt word assembly template.

[0149] In this embodiment, when there are no invalid results in each group of prompt words under the teaching syllabus, subsequent steps can be performed. When there are invalid results in each group of prompt words under the teaching syllabus, the invalid results are adjusted, which can be re-matching knowledge point information according to the subject knowledge point repository, adjusting the number of words, etc.

[0150] In this embodiment, when it is determined that there is at least one group of prompt words without a line number range in each group of prompt words under the teaching outline, all prompt words without a line number range are deleted.

[0151] There is a certain probability that the data structure of the results of the large model will be biased, so it is necessary to use the large model result parsing module to extract the valid knowledge point information and row number information contained in the results.

[0152] In the text information returned by the large model, if the format of a single line of text does not meet the following format:

[0153] <<[Line H:1-H:1] TITLE: Outline title within 30 words - INFO: Outline summary, within 50 to 150 words - KNOW: Knowledge points included in the outline. >>.

[0154] This is an invalid result. Since the invalid result still contains an overview of the course content, the knowledge points and line number information must be extracted in stages. When the result fully meets the format, the regular matching technology is used to extract the knowledge points and line number information. If the result does not fully meet the format, the keyword search technology is used to search for the H, title, INFO, and KNOW keywords to obtain the text information after the keywords and fill in the knowledge point information and line number information to the maximum extent. If the returned result has no line number information, the record is discarded.

[0155] In this embodiment, a large model result parsing module is used to extract valid knowledge point information and line number information contained in the results. For single-line text that does not meet the format, the keyword search technology is used to obtain the text information after the keyword, and the knowledge point information and line number information are supplemented to the maximum extent. If there is no line number information, the record is discarded.

[0156] The specific process of the knowledge point annotation scheme of the video resource of the present invention is as follows: Figure 2As shown, the recording and broadcasting system is equipped with a video extraction audio module for extracting audio files of video resources to be annotated, an ASR module for voice recognition of audio files, a Prompt assembly engine for prompt word assembly, and a large model introduces the output of the Prompt assembly engine and the subject knowledge point repository to obtain a teaching outline. A large model result parsing module is provided to extract valid knowledge point information and line number range contained in the result, and a line number timestamp mapping module is provided to determine the second timestamp information corresponding to each group of prompt words under the teaching outline according to the first mapping data table, and save the final index information.

[0157] The present invention provides a solution for extracting knowledge points based on a large model and automatically marking the knowledge points of recorded course resources. First, the audio information of the video file is obtained and extracted as the text of the teaching content. The above text is merged with the subject knowledge point repository to construct the large model prompt words of the subject, so that the large model can automatically complete the analysis of the knowledge points and corresponding sentences of the teaching. The analysis results of the large model are converted, the timestamps of the knowledge points in the video are marked, and an index of the association between the video and the knowledge points is constructed. In the recording and broadcasting service, the user uses the index to query the teaching video corresponding to the knowledge point, and can accurately locate the playback position according to the timestamp of the knowledge point in the video. The video retrieval efficiency when the user uses the recording and broadcasting system is effectively improved. The recording and broadcasting system extracts the teaching text information in the video file, and then uses the knowledge reasoning ability of the large model to analyze the knowledge points described in the video and the time position of the knowledge points.

[0158] It should be noted that the contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

[0159] In the present embodiment, a knowledge point annotation system for video resources is also provided. A single system is used to implement the above-mentioned embodiments and optional implementation methods. The descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0160] Figure 3 It is a structural diagram of a knowledge point annotation system for video resources according to an embodiment of the present invention.

[0161] The present invention provides a knowledge point annotation system for video resources, which is applied to a recording and broadcasting system, such as Figure 3 As shown, the knowledge point annotation system of the video resource includes:

[0162] The first processing module 11 is used to determine each sentence corresponding to the audio file of the video resource to be annotated in the recording and broadcasting system.

[0163] The second processing module 12 is used to determine the first timestamp information corresponding to each sentence, and to label the sentences in sequence to obtain the row number corresponding to each sentence. The first timestamp information corresponding to each sentence includes: a first start timestamp and a first end timestamp corresponding to each sentence. The first mapping data table is determined according to the correspondence between the first timestamp information corresponding to each sentence and the row number.

[0164] The third processing module 13 is used to determine the teaching outline according to each sentence and the line number corresponding to each sentence. The teaching outline includes: at least one group of prompt words, each group of prompt words includes a line number range and knowledge point information stored in a corresponding relationship, and the knowledge point information is determined by analyzing the knowledge point of each sentence.

[0165] The fourth processing module 14 is used to determine the second timestamp information corresponding to each group of prompt words under the teaching outline according to the first mapping data table. The second timestamp information corresponding to each group of prompt words includes: a second start timestamp and a second end timestamp corresponding to each group of prompt words. The second start timestamp is the first start timestamp corresponding to the smallest row number in the row number range, and the second end timestamp is the first end timestamp corresponding to the largest row number in the row number range.

[0166] The fifth processing module 15 is used to construct index information according to the second timestamp information corresponding to each group of prompt words and the knowledge point information in each group of prompt words. A single index information includes the second timestamp information corresponding to a single group of prompt words and the knowledge point information in the group of prompt words.

[0167] The sixth processing module 16 is used to label the video resource to be labeled according to the index information.

[0168] In an optional implementation, the first processing module is specifically used to determine the audio file of the video resource to be annotated in the recording and broadcasting system, determine the text information corresponding to the audio file, and split the text information according to preset sentence separators to obtain individual sentences.

[0169] In an optional implementation, the third processing module is specifically used to pre-build a subject knowledge point repository based on the knowledge point information in each subject textbook. The subject knowledge point repository includes the knowledge point information in the subject textbook corresponding to the video resource to be annotated.

[0170] Each sentence is analyzed to obtain initial prompt words. A single group of initial prompt words includes an initial line number range and initial knowledge point information stored in a corresponding relationship.

[0171] The initial prompt words are written into a preset prompt word assembly template to obtain an initial outline, and the initial outline is adjusted according to the subject knowledge point repository to obtain a teaching outline.

[0172] In an optional implementation, the third processing module is further configured to determine the teaching outline according to each sentence and the line number corresponding to each sentence through a pre-built large model.

[0173] In an optional implementation, the knowledge point annotation system for video resources further includes: a verification module for determining whether there are invalid results in each group of prompt words under the teaching outline after determining the teaching outline according to each sentence and the line number corresponding to each sentence. Invalid results indicate that the preset prompt word assembly template is not satisfied and includes the line number range.

[0174] When there are invalid results in each group of prompt words under the teaching outline, the invalid results are adjusted until the invalid results meet the preset prompt word assembly template.

[0175] In an optional implementation, the line numbers in the line number range corresponding to each group of prompt words in the teaching outline are consecutive numbers.

[0176] In an optional implementation, the knowledge point annotation system for video resources further includes: a saving module, which is used to set a saving interface in the recording and broadcasting system. The saving interface is used to save all index information of the video resource to be annotated.

[0177] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0178] The knowledge point annotation system of the video resource in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0179] The embodiment of the present invention also provides a computer device having the above Figure 3 The knowledge point annotation system of the video resource shown.

[0180] See also Figure 4 , Figure 4 Schematic diagram of the hardware structure of the computer device according to the embodiment of the present invention. Figure 4As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In an optional embodiment, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor device). Figure 4 A processor 10 is taken as an example.

[0181] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0182] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0183] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating device, an application required for at least one function. The data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage devices. In an optional embodiment, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0184] The memory 20 may include a volatile memory, such as a random access memory. The memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive. The memory 20 may also include a combination of the above-mentioned types of memory.

[0185] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0186] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc. Further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0187] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0188] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A knowledge point annotation method for video resources, applied to a recording and broadcasting system, characterized in that: include: Determine each sentence corresponding to the audio file of the video resource to be annotated in the recording and broadcasting system; Determine the first timestamp information corresponding to each sentence, and label them in sequence according to the sentence order to obtain the line number corresponding to each sentence; The first timestamp information corresponding to each sentence includes: a first start timestamp and a first end timestamp corresponding to each sentence; Determine a first mapping data table according to the correspondence between the first timestamp information corresponding to each clause and the row number; Determine a teaching outline according to each sentence and the line number corresponding to each sentence; the teaching outline includes: at least one group of prompt words, each group of prompt words includes a line number range and knowledge point information stored in a corresponding relationship, and the knowledge point information is determined by analyzing the knowledge points of each sentence; Determine the second timestamp information corresponding to each group of prompt words under the teaching outline according to the first mapping data table; the second timestamp information corresponding to each group of prompt words includes: a second start timestamp and a second end timestamp corresponding to each group of prompt words; the second start timestamp is the first start timestamp corresponding to the smallest row number in the row number range, and the second end timestamp is the first end timestamp corresponding to the largest row number in the row number range; Constructing index information according to the second timestamp information corresponding to each group of prompt words and the knowledge point information in each group of prompt words; a single piece of index information includes the second timestamp information corresponding to a single group of prompt words and the knowledge point information in the group of prompt words; The video resource to be marked is marked according to the index information.

2. The method according to claim 1, characterized in that: The step of determining each sentence corresponding to the audio file of the video resource to be annotated in the recording and broadcasting system includes: An audio file of a video resource to be annotated in a recording and broadcasting system is determined, text information corresponding to the audio file is determined, and the text information is split according to a preset sentence separator to obtain individual sentences.

3. The method according to claim 1, characterized in that Determining the teaching outline according to each sentence and the line number corresponding to each sentence includes: Pre-build a subject knowledge point repository based on the knowledge point information in each subject textbook; the subject knowledge point repository includes the knowledge point information in the subject textbook corresponding to the video resource to be annotated; Analyze each sentence to obtain initial prompt words; a single group of initial prompt words includes an initial line number range and initial knowledge point information stored in a corresponding relationship; The initial prompt words are written into a preset prompt word assembly template to obtain an initial outline, and the initial outline is adjusted according to the subject knowledge point repository to obtain the teaching outline.

4. The method according to claim 3, characterized in that: Determining the teaching outline according to each sentence and the line number corresponding to each sentence includes: The teaching outline is determined by a pre-built large model based on each sentence and the line number corresponding to each sentence.

5. The method according to claim 3, characterized in that: After determining the syllabus based on the clauses and the line numbers corresponding to the clauses, it also includes: Determine whether there are invalid results in each group of prompt words under the teaching outline; the invalid results indicate that the preset prompt word assembly template is not satisfied and includes a line number range; When invalid results exist in each group of prompt words under the teaching outline, the invalid results are adjusted until the invalid results meet the preset prompt word assembly template.

6. The method according to claim 1, characterized in that The line numbers in the line number range corresponding to each group of prompt words under the teaching outline are consecutive numbers.

7. The method according to claim 1, characterized in that The method further comprises: A saving interface is set in the recording and broadcasting system; the saving interface is used to save all index information of the video resource to be annotated.

8. A knowledge point annotation system for video resources, applied to a recording and broadcasting system, characterized in that: include: The first processing module is used to determine each sentence corresponding to the audio file of the video resource to be annotated in the recording and broadcasting system; The second processing module is used to determine the first timestamp information corresponding to each sentence, and label the sentences in sequence to obtain the line number corresponding to each sentence; The first timestamp information corresponding to each sentence includes: a first start timestamp and a first end timestamp corresponding to each sentence; determining a first mapping data table according to the correspondence between the first timestamp information corresponding to each sentence and the line number; The third processing module is used to determine the teaching outline according to each sentence and the line number corresponding to each sentence; the teaching outline includes: at least one group of prompt words, each group of prompt words includes a line number range and knowledge point information stored in a corresponding relationship, and the knowledge point information is determined by analyzing the knowledge points of each sentence; A fourth processing module is used to determine the second timestamp information corresponding to each group of prompt words under the teaching outline according to the first mapping data table; the second timestamp information corresponding to each group of prompt words includes: a second start timestamp and a second end timestamp corresponding to each group of prompt words; the second start timestamp is the first start timestamp corresponding to the smallest row number in the row number range, and the second end timestamp is the first end timestamp corresponding to the largest row number in the row number range; A fifth processing module, configured to construct index information according to the second timestamp information corresponding to each group of prompt words and the knowledge point information in each group of prompt words; a single piece of index information includes the second timestamp information corresponding to a single group of prompt words and the knowledge point information in the group of prompt words; The sixth processing module is used to mark the video resources to be marked according to the index information.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the knowledge point labeling method for video resources according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the knowledge point labeling method for video resources according to any one of claims 1 to 7.

Images