Method for intelligently extracting copyright information based on irregular dynamic subtitles

Abstract

The invention discloses a method for intelligently extracting copyright information based on irregular dynamic subtitles. The method comprises the following steps: S1, inputting a program video sequence; s2, carrying out OCR (Optical Character Recognition) on the input program video sequence; s3, extracting copyright roles and entities in the video sequence; and S4, constructing a dynamic subtitle relation model, and generating a copyright information table. According to the method, technical means such as video text recognition, role extraction, role automatic verification and completion, entity extraction, entity knowledge enhancement verification and relation model matching based on dynamic subtitle multi-frame image fusion are adopted; according to the invention, accurate and automatic extraction of copyright information of directors, scripts, plans, words, compositions, singing and the like in program videos containing irregular subtitles is realized. According to the method, manual intervention is not needed, the copyright information extraction efficiency and accuracy are improved, and powerful support is provided for legal right protection of a video content creator.

Description

Technical Field

[0001] The invention relates to the technical field of video editing, and in particular to a method for intelligently extracting copyright information based on irregular dynamic subtitles. Background Art

[0002] With the rapid development and popularization of Internet video technology, video content has become an indispensable part of people's daily life. However, with the sharp increase in the number of videos, the issue of content copyright protection has become increasingly prominent, copyright disputes have occurred frequently, and the accurate extraction and efficient management of copyright information have become key issues that need to be solved urgently. At the video program level, the cataloging of copyright information is a basic link in the copyright protection system, which is of great significance for safeguarding the legitimate rights and interests of video content creators. However, at present, the cataloging of video programs mostly relies on manual work, which is not only time-consuming and labor-intensive, but also prone to errors and difficult to meet the needs of large-scale video content management. Although conventional OCR (optical character recognition) technology can recognize text information in videos, since copyright information is often embedded in the prologue of program videos in the form of irregular subtitles, conventional OCR technology is difficult to accurately extract and recognize this information, and manual cataloging is still required. Therefore, it is particularly important to develop a method that can intelligently extract copyright information from irregular subtitles of program videos. Summary of the invention

[0003] In order to solve the above problems, the present invention provides a method for intelligently extracting copyright information based on irregular dynamic subtitles, comprising the following steps: S1. inputting a program video sequence; S2. performing OCR recognition on the input program video sequence; S3. extracting copyright roles and entities in the video sequence; S4. constructing a dynamic subtitle relationship model and generating a copyright information table.

[0004] Furthermore, the S2 step specifically includes the following sub-steps: S21. Select multiple frames from the program video sequence; S22. Perform OCR recognition on each selected frame and convert the characters in the frame into readable text information; S23. Fuse the OCR recognition structure; S24. Verify and complete the fused result; S25. Output the verified and completed OCR result.

[0005] Furthermore, the multiple frames selected in step S21 are specifically frames containing subtitles and having a similarity of not less than 90%.

[0006] Furthermore, the step S24 specifically includes the following sub-steps: S241. Checking errors in single-frame recognition; S242. Completing missing or erroneous characters; wherein the errors checked in step S241 specifically include: character misrecognition, character overflow, and character position error.

[0007] Furthermore, the S3 step specifically includes the following sub-steps: S31. Constructing a large language model, a copyright information role knowledge base and an external knowledge base; S32. Extracting copyright roles and entity information respectively; S33. Verifying and completing the extracted copyright roles and entity information with the data information in the copyright information role knowledge base and the external knowledge base respectively.

[0008] Furthermore, the extraction of copyright roles in the step S32 specifically includes the following steps: A1. Inputting the extracted OCR text results into the large language model; A2. Setting specific copyright role tags to filter out the roles that best match these tags, and calculating the role matching confidence; A3. Setting a confidence threshold, and comparing the role matching confidence with the confidence threshold; wherein the specific copyright role tags in step A2 specifically include: director, screenwriter, lyricist, composer, planner, and singer.

[0009] Furthermore, the extraction of entity information in the S32 step specifically includes the following steps: B1. Inputting the extracted OCR text results into the large language model; B2. Setting specific entity tags and filtering according to the entity tags; B3. Confidence scoring of the recognition results; B4. Setting a confidence threshold and comparing the entity confidence score with the confidence threshold; wherein the entity tags specifically include: person name, organization name, publication time.

[0010] Furthermore, the step S4 specifically includes the following sub-steps: S41. Constructing a dynamic subtitle relationship model; S42. Constructing a scheduling pattern in the subtitles based on the identified copyright role information and entity information; S43. Matching the scheduling pattern with the dynamic subtitle relationship model; S44. Generating a final copyright information table based on the matching results.

[0011] Furthermore, the dynamic subtitle relationships stored in the dynamic subtitle relationship model specifically include: horizontal left alignment, horizontal centering, and vertical left alignment.

[0012] The present invention provides a method for intelligently extracting copyright information based on irregular dynamic subtitles, which has the following beneficial effects: The present invention realizes accurate and automatic extraction of copyright information such as director, screenwriter, planner, lyricist, composer, singer, etc. in program videos with irregular subtitles through technical means such as video text recognition, role extraction, automatic role verification and completion, entity extraction, entity knowledge enhanced verification and relationship model matching based on dynamic subtitle multi-frame fusion. This method does not require manual intervention, improves the efficiency and accuracy of copyright information extraction, and provides strong support for the protection of the legitimate rights and interests of video content creators. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0014] Figure 1 The present invention provides a flow chart of the method. DETAILED DESCRIPTION

[0015] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0016] The following is a detailed description of the implementation method of the present invention in conjunction with the accompanying drawings. Only some embodiments are described, not all embodiments. For the purpose of clarity, representations and descriptions that are not related to the present invention are omitted in the drawings and descriptions.

[0017] In order to have a clearer understanding of the technical features, purposes and beneficial effects of the present invention, the technical solution of the present invention is now described in detail below. Obviously, the implementation cases described are part of the embodiments of the present invention, not all of the embodiments, and cannot be understood as limiting the scope of the implementation of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection scope of the present invention.

[0018] like Figure 1 As shown, the present invention provides a method for intelligently extracting copyright information based on irregular dynamic subtitles, comprising the following steps: S1. inputting a program video sequence; S2. performing OCR recognition on the input program video sequence; S3. extracting copyright roles and entities in the video sequence; S4. constructing a dynamic subtitle relationship model and generating a copyright information table.

[0019] Step S2 specifically includes the following sub-steps: S21. Select multiple frames from the program video sequence; S22. Perform independent OCR recognition on each selected frame and convert the characters in the frame into readable text information; S23. Fuse the OCR recognition structure; S24. Verify and complete the fused result; S25. Output the verified and completed OCR result. The multiple frames selected in step S21 are specifically: frames containing subtitles and with a similarity of not less than 90%.

[0020] Step S24 specifically includes the following sub-steps: S241. Checking errors in single-frame recognition; S242. Completing missing or incorrect characters; wherein the errors checked in step S241 specifically include: character misrecognition, character overflow, and character position error.

[0021] Step S3 specifically includes the following sub-steps: S31. Construct a large language model, a copyright information role knowledge base and an external knowledge base; S32. Extract copyright roles and entity information respectively; S33. Verify and complete the extracted copyright roles and entity information with the data information in the copyright information role knowledge base and the external knowledge base respectively.

[0022] Extracting copyright roles specifically includes the following steps: A1. Inputting the extracted OCR text results into the large language model; A2. Setting specific copyright role tags to filter out the roles that best match these tags, and calculating the role matching confidence; A3. Setting the confidence threshold, and comparing the role matching confidence with the confidence threshold; wherein the specific copyright role tags described in step A2 specifically include: director, screenwriter, lyricist, composer, planner, and singer.

[0023] Establish a knowledge base of copyright information roles, including directors, screenwriters, planners, lyricists, composers, singers, etc. Compare the automatically extracted roles with the standard roles in the copyright information role knowledge base to check whether there are any missing or incorrect roles. If there are missing roles, reason based on the relationship chain in the knowledge graph and try to extract or infer the missing roles from the text. If there are incorrect roles, correct them based on the correct role information in the knowledge graph. At the same time, use the weight information in the knowledge graph to sort the extracted roles to ensure that the most important roles are extracted and verified first.

[0024] Extracting entity information specifically includes the following steps: B1. Inputting the extracted OCR text results into the large language model; B2. Setting specific entity tags and filtering according to the entity tags; B3. Confidence scoring of recognition results; B4. Setting a confidence threshold and comparing the entity confidence score with the confidence threshold: retaining results greater than the threshold to ensure that the extracted entity names have high accuracy; wherein the entity tags specifically include: person names, organization names, and publication times.

[0025] Take the following text information "Producer--XXX Performing Arts Company, Director--Zhang San, Publisher--May 2024" as an example: First, the system sets a threshold, such as 90%. Then, after inputting the example text into the pre-trained large model, the entities and corresponding confidence scores are extracted: XXX Entertainment Company (confidence score 92%), Zhang San (confidence score 98%), May 2024 (confidence score 69%). Finally, each entity and its confidence score are compared with the system threshold (90%), and the results with a confidence score of more than 90% are retained, namely XXX Entertainment Company and Zhang San.

[0026] Step S4 specifically includes the following sub-steps: S41. Constructing a dynamic subtitle relationship model; S42. Constructing a scheduling pattern in the subtitles based on the identified copyright role information and entity information; S43. Matching the scheduling pattern with the dynamic subtitle relationship model; S44. Generating a final copyright information table based on the matching results. The dynamic subtitle relationships stored in the dynamic subtitle relationship model specifically include: horizontal left alignment, horizontal center alignment, and vertical left alignment.

[0027] Establish a dynamic subtitle relationship model, such as horizontal left alignment, horizontal centering, vertical left alignment, etc. Based on the video OCR recognition results of multi-frame fusion and the extracted role and entity name information, construct the typesetting pattern in the subtitles. Match the typesetting pattern with the dynamic subtitle relationship model to find the most suitable typesetting format and position relationship. Finally, according to the matching results, match the roles and entity names one by one to generate the final copyright information table.

[0028] The present invention realizes accurate extraction of copyright information such as directors, screenwriters, planners, lyricists, composers, and singers through OCR recognition, role extraction, automatic role completion, entity extraction, entity knowledge enhancement verification, and relationship model matching without human intervention.

[0029] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A method for intelligently extracting copyright information based on irregular dynamic subtitles, characterized in that: The following steps are involved: S1. Input program video sequence; S2. Perform OCR recognition on the input program video sequence; S3. Extract copyright roles and entities in video sequences; S4. Build a dynamic subtitle relationship model and generate a copyright information table.

2. The method for intelligently extracting copyright information based on irregular dynamic subtitles according to claim 1 is characterized in that: The S2 step specifically includes the following sub-steps: S21. Select multiple frames from the program video sequence; S22 performs OCR recognition on each selected frame and converts the characters in the picture into readable text information; S23. Fusing the OCR recognition structure; S24. Verify and complete the fused result; S25. Output the verified and completed OCR result.

3. The method for intelligently extracting copyright information based on irregular dynamic subtitles according to claim 2 is characterized in that: The multiple frames selected in step S21 are specifically frames containing subtitles and having a similarity of not less than 90%.

4. The method for intelligently extracting copyright information based on irregular dynamic subtitles according to claim 2 is characterized in that: The step S24 specifically includes the following sub-steps: S241. Checking errors in single frame recognition; S242. Complete missing or incorrect characters; The errors detected in step S241 specifically include: character misrecognition, character overflow, and character position error.

5. The method for intelligently extracting copyright information based on irregular dynamic subtitles according to claim 1 is characterized in that: The S3 step specifically includes the following sub-steps: S31. Build a large language model, copyright information role knowledge base and external knowledge base; S32. Extracting copyright roles and entity information respectively; S33. Verify and complete the extracted copyright roles and entity information with the copyright information role knowledge base and the data information in the external knowledge base respectively.

6. The method for intelligently extracting copyright information based on irregular dynamic subtitles according to claim 5 is characterized in that: The step S32 of extracting the copyright role specifically includes the following steps: A1. Input the extracted OCR text results into the large language model; A2. Set specific copyright role tags to filter out roles that best match these tags, and calculate role matching confidence; A3. Set a confidence threshold and compare the role matching confidence with the confidence threshold; The specific copyright role labels described in step A2 specifically include: director, screenwriter, lyricist, composer, planner, and singer.

7. The method for intelligently extracting copyright information based on irregular dynamic subtitles according to claim 5 is characterized in that: The extraction of entity information in step S32 specifically includes the following steps: B1. Input the extracted OCR text results into the large language model; B2. Set specific entity tags and filter by entity tags; B3. Confidence scoring of recognition results; B4. Set a confidence threshold and compare the entity confidence score with the confidence threshold; The entity tags specifically include: person name, organization name, and publication time.

8. The method for intelligently extracting copyright information based on irregular dynamic subtitles according to claim 1 is characterized in that: The S4 step specifically includes the following sub-steps: S41. Construct dynamic subtitle relationship model; S42. Constructing a scheduling pattern in the subtitles based on the identified copyright role information and entity information; S43. Matching the shift pattern with the dynamic subtitle relationship model; S44. Generate a final copyright information table based on the matching results.

9. The method for intelligently extracting copyright information based on irregular dynamic subtitles according to claim 8 is characterized in that: The dynamic subtitle relationships stored in the dynamic subtitle relationship model specifically include: horizontal left alignment, horizontal centering, and vertical left alignment.

Images