Course subtitle correction method and terminal device

Abstract

This invention discloses a course subtitle correction method and terminal device, relating to the field of data processing technology. It eliminates the need for manual keyword pre-setting, automatically identifying and correcting subtitles, replacing the tedious work of manual word-by-word correction and significantly reducing the basic cost of subtitle production. It is suitable for courses with high terminology density and boasts high accuracy. The key points of the solution are: extracting audio information from the course video and converting it into original course subtitles; acquiring the teaching materials corresponding to the course video and converting each page of the teaching materials into images, extracting text information from each image to form a course knowledge base; segmenting the original course subtitles into multiple text fragments according to semantic logic or a preset fixed length, and correcting each text fragment based on a large language model and the course knowledge base; and sequentially concatenating the corrected text fragments to obtain the corrected course subtitles.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and terminal device for correcting course subtitles. Background Technology

[0002] With the rapid development of educational informatization, online courses, pre-recorded lectures, and other digital teaching methods have become one of the mainstream education models. Course subtitles, as an important supplement to teaching content, help learners accurately capture key teaching points and aid in understanding professional knowledge. Their accuracy and standardization directly affect teaching quality and learning experience. Course subtitle correction is a process for automatically generated subtitles from teaching videos and audio. Through semantic understanding, terminology standardization, manual correction, and timing alignment, it corrects and optimizes subtitles for typos, grammatical errors, recognition errors, deviations in professional terminology, and timing shifts, ensuring that the subtitle content is accurate, fluent, and standardized, and consistent with the teaching materials and the teacher's lecture content.

[0003] Current course subtitle correction mainly relies on automatic speech recognition technology to convert speech to text and then automatically generate course subtitles. That is, automatic speech recognition technology is used to process the teaching audio into text and automatically generate corresponding course subtitles. After the course subtitles are generated, they are corrected manually.

[0004] While the aforementioned existing technologies possess a certain degree of accuracy in general scenarios, they require extensive manual calibration, consuming significant time and resources. This is particularly true in the field of medical and nursing training and education, where the numerous complex, lengthy, and similar technical terms, traditional Chinese characters for medicinal herbs, and obscure characters in medicine and pharmacy impose obvious limitations on the existing technologies, resulting in low accuracy. Summary of the Invention

[0005] This invention provides a course subtitle correction method and terminal device, which automatically identifies and corrects keywords without the need for manual keyword pre-setting, replacing the tedious work of manual word-by-word correction and significantly reducing the basic cost of subtitle production; it is suitable for courses with high terminology density and has a high correction accuracy rate.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of this invention provides a course subtitle correction method, comprising: Extract the audio information from the course video and convert the audio information into the original course subtitles; Obtain the teaching materials corresponding to the course videos, convert the teaching materials into images page by page, and extract the text information from each image to form a course knowledge base; The original course subtitles are divided into multiple text segments according to semantic logic or a preset fixed length, and each text segment is corrected based on a large language model and the course knowledge base; The corrected text segments are then concatenated in sequence to obtain the course correction subtitles.

[0007] Furthermore, after sequentially concatenating each corrected text segment to obtain the course correction subtitle, the process also includes: The training sample set is determined based on preset information; Construct a subtitle fine-tuning model based on the training sample set; The course correction subtitles are adjusted according to the subtitle fine-tuning model to obtain the course adjusted subtitles.

[0008] Furthermore, it also includes: determining a verification set and a test set based on the preset information; Accordingly, the subtitle fine-tuning model is validated based on the validation set; The subtitle fine-tuning model is evaluated based on the test set.

[0009] Furthermore, a subtitle fine-tuning model is constructed based on the training sample set, including: The training sample set is input into a large language model, and optimized using domain-adaptive fine-tuning technology to obtain model fine-tuning parameters; A subtitle fine-tuning model is constructed based on the model's fine-tuning parameters.

[0010] Furthermore, the preset information includes at least two of the following: speaker identifier, course identifier, course domain label, audio information, teaching materials, and course correction subtitles.

[0011] Furthermore, it also includes: Determine the speaker's language expression characteristics; The subtitle fine-tuning model is trained based on the language expression characteristics. Training stops when the data indicators in the validation set meet the preset requirements, and the trained subtitle fine-tuning model is obtained.

[0012] Furthermore, after adjusting the course correction subtitles according to the subtitle fine-tuning model, the process also includes: Compare the course correction captions and the course adjustment captions, identify the differences between them, and display them.

[0013] Furthermore, before extracting the audio information from the course videos, the process also includes: Determine whether there are any manual correction records in the course video. If so, use the manual correction records as the course correction subtitles for the corresponding parts.

[0014] Furthermore, the difference information includes: The difference information includes one or more of the following: terminology adjustment, expression optimization, grammar correction, format adjustment, and typo correction.

[0015] A second aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the course subtitle correction method described above.

[0016] This invention provides a course subtitle correction method and terminal device. First, the audio information of the course video is converted into original course subtitles. Then, the text information from the corresponding teaching materials is extracted to form a course knowledge base. Next, the course subtitles are segmented into multiple text fragments according to semantic logic or a preset fixed length. Each text fragment is corrected based on a large language model and the course knowledge base. Finally, the corrected text fragments are sequentially spliced ​​together to obtain the corrected course subtitles. Compared with existing technologies, this invention eliminates the need for manually preset keywords, automatically identifying and correcting them, replacing the tedious work of manual word-by-word correction, and significantly reducing the basic cost of subtitle production. It is suitable for courses with high terminology density and has a high correction accuracy. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. The drawings are only used to illustrate the implementation methods and are not intended to limit the present invention.

[0018] Figure 1 This is a schematic diagram of a course subtitle correction method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of another course subtitle correction method in an embodiment of the present invention; Figure 3 This is a schematic diagram of the composition structure of a course subtitle correction terminal device in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art; the terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the terms “comprising” and “having” and any variations thereof in the specification, claims and foregoing drawings, but is intended to cover non-exclusive inclusion.

[0021] In the description of the embodiments of this invention, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this invention, "multiple" means two or more, unless otherwise explicitly defined.

[0022] In the description of the embodiments of this invention, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0023] In the description of the embodiments of the present invention, the term "multiple" refers to two or more (including two), similarly, "multiple groups" refers to two or more (including two groups), and "multiple pieces" refers to two or more (including two pieces).

[0024] In the description of the embodiments of the present invention, the technical terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of the present invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of the present invention.

[0025] In the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of the present invention can be understood according to the specific circumstances.

[0026] Example 1

[0027] This invention provides a method for correcting course subtitles, such as... Figure 1 As shown, it includes: S101. Extract the audio information from the course video and convert the audio information into the original course subtitles.

[0028] Course videos refer to teaching videos that require subtitle correction, are used in teaching scenarios (such as online courses and teaching explanations), and whose subtitles have issues that need to be corrected (such as ASR recognition errors, non-standard terminology, timing deviations, etc.). These videos need to be processed through semantic error correction, standardization optimization, and other methods to ensure that the subtitle content is consistent with the course teaching content and meets teaching needs.

[0029] There are two main ways to obtain course videos: the first is manual upload, where users (such as course administrators or instructors) manually upload the course video files to be corrected through the system's preset upload interface; the second is automatic retrieval by the system: if the system has already stored course-related video resources, it can automatically retrieve the course videos to be corrected by reading the system's internal course video storage path.

[0030] Audio information refers to the identifiable and processable sound signal data extracted from course videos, specifically the digital information corresponding to the speech signals generated during the teacher's lecture. Original course subtitles refer to the text information generated by converting the audio information in the course videos using speech recognition technology; these are the initial, uncorrected textual carriers used to assist learners in understanding the teaching content.

[0031] S102. Obtain the teaching materials corresponding to the above course videos, convert the teaching materials into images page by page, and extract the text information from each page of the images to form a course knowledge base.

[0032] Teaching courseware refers to all kinds of teaching materials produced to assist in course teaching and to convey teaching knowledge. It includes, but is not limited to, PPT courseware, WORD, PDF, OFD, and txt format files. It is used to assist teachers in teaching and to help students understand the teaching content. It is highly matched with the course teaching theme and is an important part of course teaching.

[0033] There are two main ways to obtain teaching materials: The first is manual uploading, where users manually upload teaching materials (usually in PPT, PDF, or other formats) corresponding to the course videos to be corrected through the system's preset upload interface; the second is automatic retrieval by the system, which automatically retrieves the corresponding teaching materials using course ID, video ID, and other associated information. Text information is extracted from each page's images to form a course knowledge base. Here, "text" specifically refers to all readable character information extracted from the teaching material images using optical character recognition technology, including course titles, knowledge points, professional terminology, and main text. The course knowledge base is a set of authoritative reference data specific to the course, formed by organizing, classifying, and structuring the extracted text information according to the course chapter logic. It is used for subsequent subtitle semantic correction and includes keyword lists, professional terminology, and semantic understanding content.

[0034] S103. Divide the course subtitles into multiple text segments according to semantic logic or a preset fixed length, and correct each text segment based on a large language model and course knowledge base.

[0035] Semantic logic refers to the semantic integrity and logical coherence expressed by language, that is, the meaning conveyed by a sentence or text fragment is coherent and complete, conforms to people's daily expression habits and thinking logic, and can clearly reflect a complete meaning, rather than scattered or fragmented words or short sentences.

[0036] Preset fixed length refers to a pre-set, uniform text length standard (such as 30-50 characters per segment or 10 lines per segment) when segmenting course subtitles. This serves as a fixed basis for splitting course subtitle segments without the need to combine semantic logic. The text is split according to the set uniform length to ensure that the length of each segment is consistent.

[0037] A text fragment refers to an independent text unit that has a relatively complete meaning or a clear scope, formed by splitting complete text content (such as course subtitles) according to certain rules (semantic logic or preset fixed length).

[0038] Large language models are artificial intelligence models trained on large-scale text data and possess capabilities such as natural language understanding, semantic analysis, grammatical error correction, and content optimization. They can receive input text / voice information and, combined with their own trained knowledge, complete operations such as semantic judgment, error correction, and content standardization.

[0039] Correction specifically refers to the process of semantic analysis and error correction of text fragments based on preset reference standards (such as course knowledge bases and professional terminology specifications).

[0040] S104. Concatenate each corrected text segment in sequence to obtain the course correction subtitle.

[0041] Among them, splicing refers to the process of combining and connecting the corrected text segments in the course subtitle correction process according to their original time sequence and semantic logic to form complete, coherent, non-repetitive and non-omitted subtitle content, while retaining the original timestamps of each text segment to ensure the semantic coherence and temporal synchronization of the spliced ​​subtitles.

[0042] This invention provides a course subtitle correction method. First, the audio information of the course video is converted into the original course subtitles. Then, the text information in the teaching materials corresponding to the course video is extracted to form a course knowledge base. Next, the course subtitles are divided into multiple text segments according to semantic logic or a preset fixed length. Each text segment is corrected based on a large language model and the course knowledge base. Finally, the corrected text segments are sequentially spliced ​​together to obtain the corrected course subtitles. Compared with the prior art, this invention does not require manual keyword preset, automatically identifies and corrects keywords, replacing the tedious work of manual word-by-word correction, and significantly reducing the basic cost of subtitle production. It is suitable for courses with high terminology density and has a high correction accuracy.

[0043] Example 2

[0044] This invention provides a method for correcting course subtitles, such as... Figure 2 As shown, it includes: S201. Extract the audio information from the course video and convert the audio information into the original course subtitles.

[0045] Specifically, the method for extracting audio information from course videos is as follows: using audio analysis tools, pure audio information is separated from the course videos (removing irrelevant content such as video footage and background noise), while retaining the clear voice signal of the teacher. This process filters out background noise, page turning, coughing, and other interfering sounds in the classroom environment to ensure that the extracted audio contains only the teaching voice.

[0046] The method for converting audio information into original course subtitles is as follows: The audio information is standardized into a uniform format, and automatic speech recognition technology is used to process the extracted audio information into text. The automatic speech recognition technology identifies each speech segment in the audio information and converts it into corresponding text. Simultaneously, based on pauses and speech rate, it automatically matches timestamps to generate original course subtitles with timestamps.

[0047] S202. Obtain the teaching materials corresponding to the course videos, convert each page of the teaching materials into images, and extract the text information from each page of the images to form a course knowledge base.

[0048] Specifically, the method to convert teaching courseware page by page into images is as follows: Using common office / conversion tools, open the teaching courseware (PPT, PDF format), use the tool's built-in "Export / Save As" function, select the image format (JPG, PNG, etc.), and set "Export page by page" to convert the teaching courseware into independent, clear images page by page in order of page number.

[0049] Extract text information from each page's images to form a course knowledge base: Using optical character recognition technology, visual language models, or natural language models, extract all text information from each page's images (including titles, body text, text in formulas, annotations, etc.), organize and structure the extracted text information to form a course knowledge base.

[0050] S203. Determine whether there are manual correction records in the course video. If so, use the manual correction records as the course correction subtitles for the corresponding parts.

[0051] Specifically, the method for determining whether a course video has a manual correction record is as follows: Using the course video's unique identifier (such as course ID, video ID, chapter number, etc.), the system's internal database of manual correction records is searched to determine if the course video has a historical manual correction record. It should be noted that the manual correction record here refers to the final, acceptable subtitles saved after previous manual proofreading and modification of the course video's subtitles.

[0052] If a valid manual correction record is found for the course video (i.e., the record is complete, not expired, and completely matches the teaching content and chapters of the current course video without any content changes), then the manual correction record will be directly used as the course correction subtitle for the corresponding part; if no manual correction record is found or the record is invalid (e.g., the content is incomplete or does not match the current course content), then the subsequent correction process will proceed normally.

[0053] S204. Divide the course subtitles into multiple text segments according to semantic logic or a preset fixed length, and correct each text segment based on a large language model and course knowledge base.

[0054] Specifically, the course subtitles are segmented using one of the following two methods: (1) Segmentation based on semantic logic: Segmentation is based on the semantic integrity of the course subtitles. For example, a complete sentence (such as "In this lesson we will learn the basic concepts of artificial intelligence") is treated as a text fragment to avoid segmentation into incomplete semantics (such as "In this lesson we will learn") after splitting.

[0055] (2) Segmentation by preset fixed length: If the subtitle sentence is too long (e.g., more than 50 characters), it will be segmented by preset fixed length (e.g., every 30-50 characters or every 10 lines) to ensure that each text segment is of appropriate length, which is convenient for the large language model to perform semantic analysis and error correction, while preserving the semantic relationship between text segments.

[0056] The correction of each text segment based on a large language model and course knowledge base specifically includes: (1)Text errors: Correct spelling mistakes, missing words, and extra words caused by automatic speech recognition (e.g., changing "artificial only" to "artificial intelligence", and "data analysis" to "data analysis").

[0057] (2)Grammar and syntax errors: Correct ungrammatical sentences, incorrect word order, and incorrect collocations (e.g., changing "We learned artificial intelligence yesterday" to "Yesterday we learned artificial intelligence").

[0058] (3)Semantic deviations: Correct expressions that do not match the course knowledge base and incorrect professional terms (e.g., misidentifying "machine learning algorithm" as "machine learning algorithm", and correcting it based on the course knowledge base).

[0059] (4)Expression norms: Correct colloquial and non-standard expressions into standard expressions that conform to course teaching (e.g., changing the teacher's colloquial expression "Let's talk about this knowledge point today" to "In this class, we will explain this knowledge point").

[0060] S205. Concatenate each corrected text segment in order to obtain the course corrected subtitles.

[0061] Specifically, concatenating text segments in order means concatenating each corrected text segment in the order of the segmentation of the original course subtitles, ensuring that the content of the concatenated course corrected subtitles is coherent, semantically complete, without repetition, omission, or poor connection issues (for example, if the segmented text segments are "In this class, we will learn" and "the basic concepts of artificial intelligence", the concatenated result is "In this class, we will learn the basic concepts of artificial intelligence").

[0062] It should be noted here that during the concatenation of text segments, the original timestamps of each text segment are retained to ensure that the appearance and disappearance times of the text in the text segment are completely synchronized with the teacher's teaching speech, avoiding the situation of subtitles being ahead or behind.

[0063] S206. Determine the training sample set according to the preset information.

[0064] The preset information is the core exclusive feature set of the course video and the core basis for determining the training sample set, specifically including: speaker identification, course identification, audio information, teaching courseware, course corrected subtitles, course field tags (such as science and engineering, liberal arts, medicine, etc.), professional fields (such as computer, mathematics, Chinese language, etc.), course level (such as beginner, intermediate, advanced), core term library, and teacher teaching style (such as rigorous academic type, popular explanation type). In specific applications, two or more can be selected Perform preprocessing on the preset information, specifically including: denoising and silent cutting of audio information, and normalization processing of text.

[0065] The training sample set is determined by selecting sample data that highly matches the professional field, terminology, and expression style of the current course based on the course information. The core components of the sample set include: (1) "original subtitle-correct subtitle" pairing data of the current course or similar courses (such as the original subtitle with automatic speech recognition error and the correct subtitle after manual correction); (2) pairing data of professional terms specific to the current course (such as the correspondence between incorrect terms and standard terms); (3) examples of subtitle segments that conform to the expression style of the current course (such as the standard expression of academic courses and the adapted expression of popular courses).

[0066] Preferably, to ensure the stability and effectiveness of subsequent subtitle fine-tuning model training, the training sample set needs to be verified to ensure that the sample data is accurate, the terminology is consistent, and the format is standardized, and invalid samples (such as incorrect pairings or samples with confusing expressions) are removed; at the same time, the training sample set is standardized to unify the sample format, text length, and annotation specifications.

[0067] S207. Construct a subtitle fine-tuning model based on the training sample set.

[0068] S2071. Input the training sample set into the large language model and optimize it using domain-adaptive fine-tuning technology to obtain the model fine-tuning parameters.

[0069] First, the large language model selects a general pre-trained model adapted to the subtitle processing scenario (such as BERT, GPT series, etc., which have natural language understanding and error correction capabilities) as the base model.

[0070] Secondly, the validation set and test set are determined based on the preset information. The training sample set, validation set and test set can be divided in a ratio (8:1:1 or 7:2:1).

[0071] Furthermore, the subtitle fine-tuning model can be validated using a validation set and evaluated using a test set. This further ensures that the model can accurately learn the terminology conventions, error correction logic, and expression style in the training sample set, ultimately outputting model fine-tuning parameters that highly match the training sample set.

[0072] S2072. Construct a subtitle fine-tuning model based on the model fine-tuning parameters.

[0073] Specifically, the model fine-tuning parameters are the core, and the architecture of general pre-trained models (such as BERT, GPT series, etc., which have natural language understanding and error correction capabilities) is integrated to form a complete subtitle fine-tuning model.

[0074] S208. Adjust the course correction subtitles according to the subtitle fine-tuning model to obtain the course adjusted subtitles.

[0075] The course correction subtitles are input into the subtitle fine-tuning model. The subtitles are first preprocessed, including removing redundant spaces, unifying punctuation marks, and splitting excessively long sentences (consistent with the S103 segmentation logic), to ensure that the subtitle format meets the model's processing requirements.

[0076] Furthermore, it is possible to identify the speaker's language expression characteristics, such as course-specific error correction rules, terminology standards, and expression style; The subtitle fine-tuning model is trained based on the characteristics of language expression. Training stops when the data indicators in the validation set meet the preset requirements, and the trained subtitle fine-tuning model is obtained.

[0077] Specifically, the trained subtitle fine-tuning model can correct minor terminology deviations (such as unrecognized variations of technical terms in general error correction); optimize expression style to ensure that subtitles are consistent with the course teaching style, such as more rigorous academic courses and easier-to-understand popular courses.

[0078] S209. Compare the course correction subtitles and the course adjustment subtitles, identify the differences between them and display them. Specifically, (1) Align the course correction subtitles and the course adjustment subtitles in chronological order to ensure that the corresponding segments of the two subtitles match one by one; at the same time, unify the format of the two subtitles (such as punctuation and spaces). (2) Use a text comparison algorithm to compare the content of the two subtitles sentence by sentence and segment by segment, automatically identify and extract all the differences between them. The differences mainly include: terminology adjustment (such as "machine semester" changed to "machine learning"); expression optimization (such as "we will talk" changed to "this lesson will explain"); grammar correction (such as word order adjustment and collocation optimization); format adjustment (such as punctuation and capitalization correction); and misspelling correction (such as missing punctuation and similar-looking characters). (3) Classify and organize the identified differences, mark them according to the difference type (terminology, expression, grammar, format, etc.), and mark the subtitle segments and timestamps corresponding to the differences to form a clear difference comparison table, clarifying the specific content and position of each adjustment, so as to facilitate subsequent manual verification and traceability.

[0079] It should be noted that the detailed description of each step in this embodiment can be found in other embodiments, and will not be repeated here.

[0080] This invention provides a method for correcting course subtitles. First, the audio information of the course video is converted into original course subtitles. Then, the text information from the corresponding teaching materials is extracted to form a course knowledge base. Next, the course subtitles are segmented into multiple text fragments according to semantic logic or a preset fixed length. Each text fragment is corrected based on a large language model and the course knowledge base. Finally, the corrected text fragments are sequentially spliced ​​together to obtain the corrected course subtitles. Compared to existing technologies, this invention eliminates the need for manually preset keywords, automatically identifying and correcting them, replacing the tedious work of manual word-by-word correction, and significantly reducing the basic cost of subtitle production. It is suitable for courses with high terminology density and has a high accuracy rate. Simultaneously, this invention determines whether the course video to be corrected has a manual correction record. If so, the manual correction record is used as the corrected course subtitles. Even if the course video to be corrected is re-recorded, edited, or partially modified, the corrected high-quality content will not be lost, achieving "one-time correction, lifelong benefit."

[0081] Furthermore, this embodiment of the invention constructs a subtitle fine-tuning model and adjusts the course correction subtitles according to the subtitle fine-tuning model to obtain course correction subtitles, thereby improving the pertinence and adaptability of subtitle correction and increasing the accuracy of subtitle correction.

[0082] Finally, this embodiment of the invention compares the course correction subtitles and the course adjustment subtitles, identifies and displays the differences between the two, which greatly reduces the workload of manual correction and improves the efficiency of review.

[0083] Example 3

[0084] This invention provides a course subtitle correction terminal device, such as... Figure 3 As shown, it includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, it implements the steps of the course subtitle correction method described above.

[0085] The processing device (e.g., central processing unit, graphics processor, etc.) 31 can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 32 or a program loaded from storage device 38 into random access memory (RAM) 33. The RAM 33 also stores various programs and data required for the operation of the electronic device. The processing device 31, ROM 32, and RAM 33 are interconnected via bus 34. An input / output (I / O) interface 35 is also connected to bus 34.

[0086] Typically, the following devices can be connected to I / O interface 35: input devices 36 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 37 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 38 including, for example, magnetic tapes, hard disks, etc.; and communication devices 39. Communication device 39 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0087] In particular, according to some embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 39, or installed from a storage device 38, or installed from a ROM 32. When the computer program is executed by the processing device 31, it performs the functions defined in the methods of some embodiments of the present invention.

[0088] This invention provides a course subtitle correction terminal device that automatically identifies and corrects subtitles without requiring manual keyword pre-setting, replacing the tedious work of manual word-by-word correction and significantly reducing the basic cost of subtitle production; it is suitable for courses with high terminology density and has a high correction accuracy rate.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention. In particular, as long as there is no structural conflict, the various technical features mentioned in the embodiments can be combined in any way. The present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for correcting course subtitles, characterized in that, include: Extract the audio information from the course video and convert the audio information into the original course subtitles; Obtain the teaching materials corresponding to the course videos, convert the teaching materials into images page by page, and extract the text information from each image to form a course knowledge base; The original course subtitles are divided into multiple text segments according to semantic logic or a preset fixed length, and each text segment is corrected based on a large language model and the course knowledge base; The corrected text segments are then concatenated in sequence to obtain the course correction subtitles.

2. The course subtitle correction method according to claim 1, characterized in that, After sequentially concatenating each of the corrected text segments to obtain the course-corrected subtitles, the following is also included: The training sample set is determined based on preset information; Construct a subtitle fine-tuning model based on the training sample set; The course correction subtitles are adjusted according to the subtitle fine-tuning model to obtain the course adjusted subtitles.

3. The course subtitle correction method according to claim 2, characterized in that, After adjusting the course correction subtitles according to the subtitle fine-tuning model to obtain the course adjusted subtitles, the process further includes: The verification set and test set are determined based on the preset information; Accordingly, the subtitle fine-tuning model is validated based on the validation set; The subtitle fine-tuning model is evaluated based on the test set.

4. The course subtitle correction method according to claim 2, characterized in that, Constructing a subtitle fine-tuning model based on the training sample set includes: The training sample set is input into a large language model, and optimized using domain-adaptive fine-tuning technology to obtain model fine-tuning parameters; A subtitle fine-tuning model is constructed based on the model's fine-tuning parameters.

5. The course subtitle correction method according to claim 2, characterized in that, The preset information includes at least two of the following: speaker identifier, course identifier, course domain label, audio information, teaching materials, and course correction subtitles.

6. The course subtitle correction method according to claim 5, characterized in that, Also includes: Determine the speaker's language expression characteristics; The subtitle fine-tuning model is trained based on the language expression characteristics. Training stops when the data indicators in the validation set meet the preset requirements, and the trained subtitle fine-tuning model is obtained.

7. The course subtitle correction method according to claim 2, characterized in that, After adjusting the course correction subtitles according to the subtitle fine-tuning model, the following steps are also included: Compare the course correction captions and the course adjustment captions, identify the differences between them, and display them.

8. The course subtitle correction method according to claim 1, characterized in that, Before extracting the audio information from the course videos, the following steps are also included: Determine whether there are any manual correction records in the course video. If so, use the manual correction records as the course correction subtitles for the corresponding parts.

9. The course subtitle correction method according to claim 7, characterized in that, The difference information includes: The difference information includes one or more of the following: terminology adjustment, expression optimization, grammar correction, format adjustment, and typo correction.

10. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the course subtitle correction method as described in any one of claims 1 to 9.

Images