Copyright auditing method, system and device, medium and product
By identifying core protection points and compressing them into high-density data during copyright review, combined with multi-channel neural networks and automated review nodes, the problems of missed detections and increased operating costs in copyright review are solved, achieving efficient and accurate copyright review.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-26
- Publication Date
- 2026-04-03
AI Technical Summary
Existing copyright review methods face the risk of missed detections and increased operating costs when faced with an explosive increase in copyright infringement cases. In particular, traditional methods are unable to effectively handle large amounts of uploaded content when system throughput and response latency drop sharply.
By identifying the core protection points of the target file, a multi-channel neural network is used to extract core features and perform fuzzy compression. After constructing a high-density key point compression, similarity matching is performed in the feature database. Combining style bias type and infringement level, the review node is automatically determined and copyright review is carried out.
It improved the accuracy and efficiency of copyright review, reduced manual processing workload, lowered operating costs, and increased system throughput and response speed while ensuring approval quality.
Smart Images
Figure CN121786541A_ABST
Abstract
Description
Technical Field
[0001] Applications generally relate to the field of computer technology, and in particular to a copyright review method, system, device, medium, or product. Background Technology
[0002] Currently, copyright infringement is experiencing explosive growth, requiring platforms to conduct copyright checks on millions of uploaded content items daily. The mainstream copyright review method primarily involves checking the similarity between the target multimedia file and the copyrighted multimedia file. This is followed by comparing files with high similarity from a copyright database to identify potentially infringing files, which are then subject to secondary manual review to ultimately determine the infringement.
[0003] However, as the business expands, traditional auditing methods have at least two major problems: First, the auditing process is too simple, making it easy to miss checks and leading to legal risks; second, if manual checkpoints are added blindly, the system throughput and response latency will drop sharply, and operating costs will rise accordingly. Summary of the Invention
[0004] In view of the above, this application provides a copyright review method, system, device, medium, and product to address some of the deficiencies mentioned in the background art.
[0005] In a first aspect, embodiments of this application provide a copyright review method, including: Identify the core protection points of the target documents to be reviewed; In a preset feature database, a list of target files whose content overlaps with the core protection point content by a proportion greater than a threshold is determined. The first review node for the target file is determined based on the target file list; Based on the style preference type and infringement level of the target file, a second review node is determined in the first review node, and the copyright review of the target file is carried out through the second review node.
[0006] In a second aspect, embodiments of this application provide a copyright review system, including: The protection point determination module is configured to: determine the core protection points of the target file to be reviewed; The file list determination module is configured to: determine a list of target files whose overlap ratio with the content of the core protection point is greater than a threshold in a preset feature database; The first review node determination module is configured to: determine the first review node of the target file based on the target file list; The copyright review module is configured to: determine a second review node in the first review node based on the style preference type and infringement level of the target file, and conduct copyright review on the target file through the second review node.
[0007] In a third aspect, embodiments of this application provide a transmission device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the copyright review method described in the first aspect above.
[0008] In a fourth aspect, embodiments of this application provide a storage medium storing computer instructions for causing the computer to execute the copyright review method described in the first aspect.
[0009] In a fifth aspect, embodiments of this application provide a computer program product, including computer program instructions, which, when executed on a computer, cause the computer to perform the copyright review method described in the first aspect. Attached Figure Description
[0010] Figure 1 A flowchart illustrating a copyright review method provided in an embodiment of this application is shown; Figure 2 This document illustrates a flowchart of a method for determining core protection points according to an embodiment of this application. Figure 3 This illustration shows a schematic diagram of a software structure for determining core protection points according to an embodiment of this application; Figure 4 This paper shows a schematic diagram of the structure of a copyright review system provided in an embodiment of this application; Figure 5 A schematic diagram of a more specific electronic device hardware structure provided in this embodiment is shown. Detailed Implementation
[0011] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the application and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present application, not the entire structure.
[0012] As described in the background section, copyright infringement is currently experiencing explosive growth, requiring platforms to conduct copyright checks on millions of uploaded content items daily. The mainstream copyright review method primarily involves checking the similarity between the target multimedia file and the copyrighted multimedia file. This is followed by comparing files with high similarity from a copyright database to identify potentially infringing files, which are then subject to secondary manual review to ultimately determine the infringement.
[0013] However, as the business expands, traditional auditing methods have at least two major problems: First, the auditing process is too simple, making it easy to miss checks and leading to legal risks; second, if manual checkpoints are added blindly, the system throughput and response latency will drop sharply, and operating costs will rise accordingly.
[0014] Figure 1 The diagram illustrates a copyright review method according to an embodiment of the present invention, which includes the following steps: S102: Identify the core protection points of the target document to be reviewed.
[0015] In some optional embodiments, the copyright review method of this disclosure can be implemented by a copyright review device. Specifically, the copyright review device first determines the header file and file extension of the target file, and then confirms the file type of the target file based on the header file and file extension. The file type of the target file can be one of a text file, an image file, a video file, or an audio file.
[0016] Figure 2 A schematic diagram of a process for determining core protection points provided in an embodiment of this application is shown.
[0017] In some alternative embodiments, such as Figure 2 As shown, step S102 further includes the following steps: S1021: Based on the file type, the target file is format-converted through the format conversion layer to obtain the converted target file.
[0018] S1022: Based on the feature extraction layer, extract the features of the converted target file to obtain the core features.
[0019] S1023: Based on the validity filtering layer, the core features are filtered through validity activation parameters to obtain core protection points.
[0020] Figure 3 This illustration shows a schematic diagram of a software structure for determining core protection points according to an embodiment of this application.
[0021] In some alternative embodiments, please refer to Figure 2 and Figure 3 As shown, according to the copyright document Type Regarding copyright documents Convert to obtain the converted copyright file. The process involves constructing and utilizing a neural network to transform the effective location of core protection points to extract the transformed copyright file. The core protection point, obtain the core protection point For core protection points Validity filtering is performed to obtain valid core protection points. .
[0022] After determining the file type, the copyright review device can identify the core protection points of the target file using a preset neural network model. This preset neural network model includes a format conversion layer, a feature extraction layer, and a validity filtering layer. Based on the file type, the format conversion layer performs format conversion on the target file to obtain the converted target file.
[0023] In some optional embodiments, the copyright document Convert to obtain the converted copyright file. The process can be specifically described as follows: by unifying the input format of copyright files through a conversion layer, various types of copyright files can be processed. Convert to a fixed multimedia format type for subsequent processing, categorized by type. Please enter the copyright file according to the table below. Convert to the target format to obtain the converted copyright file. .
[0024] In some optional embodiments, when the file type is a text file, the text content of the target file can be read and concatenated into a string, which is then saved as a .txt file. When the file type is an image file, a preset transcoding tool (ffmpeg) can be used to convert the target file to JPEG format. When the file type is a video file, a preset transcoding tool can be used to convert the target file to MP4 format. When the file type is an audio file, a preset transcoding tool can be used to convert the target file to MP3 format.
[0025] The copyright file format conversion table is as follows:
[0026] In some optional embodiments, after obtaining the converted target file, the copyright review device can extract features from the converted target file based on the feature extraction layer to obtain core features. Then, the copyright review device can filter the core features using validity activation parameters based on the validity filtering layer to obtain core protection points.
[0027] In some optional embodiments, the process of extracting features from the converted target file based on the feature extraction layer to obtain core features can be specifically as follows: the copyright review device can construct a multi-channel neural network, then determine the position code of the converted target file, then input the converted target file and the position code into the multi-channel neural network, and output the core features of the converted target file and the effective position of the core features through the multi-channel neural network.
[0028] In some optional embodiments, the aforementioned process of constructing a multi-channel neural network can be specifically described as follows: First, a first multi-channel neural network is constructed using an attention mechanism. This first multi-channel neural network takes the converted target file and the positional encoding as input, passes through residual connections and attention weighting, and outputs contextual semantic features. Then, a second multi-channel neural network is constructed using residual fully connected layers. This second multi-channel neural network takes the contextual semantic features as input, passes through residual fully connected layers for localization, and outputs core features and the effective positions of these core features.
[0029] In some optional embodiments, the process of filtering core features based on the validity filtering layer and obtaining core protection points through validity activation parameters can be understood as the copyright review device inputting core features and valid locations into the validity filtering layer. When the confidence level of the valid location is higher than the validity activation parameter, the core feature corresponding to the valid location can be determined as the core protection point.
[0030] In some alternative embodiments, as mentioned above, to reduce errors caused by precise sampling errors, fuzzy compression can be used to fuzzy compress the parts of non-core protection points, thereby avoiding increased screening errors due to factors such as sampling accuracy and transcoding errors.
[0031] The logic for performing fuzzy compression processing on any valid core protection point can be as follows:
[0032] Where max(x) represents finding the maximum value of x, min(x) represents finding the minimum value of x, cou(x,y) represents counting the number of numbers in x that are greater than or equal to y, and z(x) represents compressing x, with the compression logic as follows:
[0033] Here, avg(x) represents taking the average value of x, | represents the concatenation operation, and x_i represents taking the data with index i from x.
[0034] In some optional embodiments, after determining the core protection points of the target file to be reviewed, the copyright review method of this disclosure can further segment the core protection points according to a preset length to obtain several segmented core protection points. Then, the segmented core protection points are fuzzily compressed to obtain fuzzily compressed core protection points. After that, the fuzzily compressed core protection points are iteratively calculated to obtain high-density key point compression. Finally, the high-density key point compression is subjected to product iteration processing to obtain high-density key point compression of the target file.
[0035] In some optional embodiments, the process of performing fuzzy compression on several segmented core protection points to obtain fuzzy compressed core protection points may include: first, performing fuzzy processing on each segmented core protection point to obtain several fuzzy processed core protection points; then, performing local compression on several fuzzy processed core protection points to obtain several compressed core protection points; and finally, sorting the several compressed core protection points to obtain fuzzy compressed core protection points.
[0036] In some optional embodiments, the process of locally compressing several fuzzy core protection points to obtain several compressed core protection points can be specifically as follows: determining the average, maximum, minimum and variance of several fuzzy core protection points, and then, based on the average, maximum, minimum and variance, locally compressing several fuzzy core protection points to obtain several compressed core protection points.
[0037] In some optional embodiments, the process of performing iterative calculations on the fuzzy compressed core protection points to obtain high-density key point compression can be understood as follows: dividing the fuzzy compressed core protection points into several segments according to a first preset number of bits, taking sub-blocks of a second preset number of bits from each segment in sequence, and performing an XOR modulo operation on each sub-block to obtain the compression code of each sub-block, then concatenating the compression codes to obtain the high-density key point compression of each segment, and finally summing up the high-density key point compression of each segment to obtain the high-density key point compression of the core protection points.
[0038] In some optional embodiments, the high-density key point compression can be expressed as: ; in, This indicates high-density compression. Indicates taking The high-density compression of key points has the following calculation logic: ; in, For XOR operation, To obtain Subscript arrive The binary encoding of bmod(x,y) represents the binary encoding of x,y. press Perform the mold taking operation. This represents the modulo-mixing parameter.
[0039] In some optional embodiments, performing product iteration processing on the high-density key point compression to obtain the high-density key point compression of the target file may include performing product iteration merging on the high-density key point compression to obtain the merged high-density key point compression, and then performing concatenation operation on the merged high-density key point compression to obtain the high-density key point compression of the target file.
[0040] In some optional embodiments, after obtaining the high-density key point compression of the target file, the copyright review method in this disclosure further includes determining the high-density key point compression of each file in the target database, and constructing a feature database based on the high-density key point compression of each file.
[0041] S104: In the preset feature database, determine a list of target files whose content overlaps with the core protection point by a proportion greater than a threshold.
[0042] In some optional embodiments, determining the target file list in the preset feature database that has an overlap ratio greater than a threshold with the core protection point content may include: compressing the high-density key points of the target files and segmenting them into segments of a preset length to form a target segment set; then compressing the high-density key points of each file in the feature database and segmenting them into segments of a preset length to form a candidate segment set; then traversing the target segment set and comparing the content of each segment in the target segment set with the content of each segment in the candidate segment set to obtain the number of segments with the same content; then determining the content overlap ratio based on the number of segments and the number of segments in the target segment set and the candidate segment set; and then generating a target file list based on the target files when the content overlap ratio is greater than the threshold.
[0043] In some optional embodiments, the process of determining the content overlap ratio based on the number of segments and the number of segments in the target segment set and the candidate segment set may include, according to the first number of segments in the target segment set and the second number of segments in the candidate segment set, using the ratio of the number of segments to the first number of segments as the content overlap ratio when the first number of segments is less than the second number of segments, and using the ratio of the number of segments to the second number of segments as the content overlap ratio when the second number of segments is less than the first number of segments.
[0044] S106: Determine the first audit node of the target file based on the target file list.
[0045] In some optional embodiments, after obtaining the target file list, the first review node for each target file can be determined based on the target file list. Specifically, the jump threshold for the approval node can be determined based on the number of historical automatic approval nodes, the total number of historical approval nodes, and the number of nodes that were returned after approval for each file in the target file list. Then, the jump threshold is compared with the jump index of each review node in the preset review node list, and the comparison result is obtained. The first review node is then determined based on the comparison result.
[0046] In some optional embodiments, the process of determining the first audit node based on the comparison results may include designating any audit node as the first audit node when the jump index of any audit node is greater than or equal to the jump threshold.
[0047] The jumping threshold can be represented as:
[0048] in, This represents the total number of historical automatic approval nodes. This represents the total number of historical approval nodes. The total number of nodes that were automatically approved and subsequently returned for re-approval in the historical node process.
[0049] S108: Based on the style bias type and infringement level of the target file, determine the second review node in the first review node, and conduct copyright review on the target file through the second review node.
[0050] In some optional embodiments, after obtaining the first review node, a second review node can be determined from the first review nodes based on the style bias type and infringement level of the target file. Specifically, this includes determining the style bias type of the target file based on the target file and its description file. Then, the authorization approval reliability coefficient of the first review node is determined based on the style bias type. Next, the completeness coefficient of the first review node is determined based on the infringement level, and the automatic execution coefficient of the first review node is calculated based on the authorization approval reliability coefficient and the completeness coefficient. Then, based on the automatic execution coefficient and the resources of each service node in the server cluster, the authorization approval capability of each service node is calculated. Finally, the second review node is determined from the first review nodes based on the authorization approval reliability coefficient, the completeness coefficient, and the authorization approval capability.
[0051] In some optional embodiments, the process of performing copyright review on the target file through the second review node may include calculating the completion degree of each review node in the second review node. The node completion degree may be determined by a weighted average of basic completion degree, time efficiency completion degree, style leap completion degree, and approver level completion degree. Then, the copyright review of the target file is performed based on the completion degree, and the final review result is obtained.
[0052] In some optional embodiments, copyright review of the target file based on completion level to obtain a final review result may include: when the completion level is greater than or equal to a preset approval threshold, performing copyright review on the target file based on the target server to obtain a first review result; when the completion level is less than the approval threshold, performing copyright review on the target file manually to obtain a second copyright review result; and generating a final review result for the target file based on the first and second review results.
[0053] In some alternative embodiments, based on style preference type Infringement level Calculate the reliability coefficient of authorization and approval. Completeness coefficient Automatic execution coefficient
[0054] In some optional embodiments, the authorization approval reliability coefficient can also be calculated using the following method. :
[0055] in, This indicates the total number of approvals at this approval node throughout history. This approval node indicates a style preference and type. The total number of approvals in history.
[0056] In some optional embodiments, the completeness coefficient can also be calculated using the following method. :
[0057] Where sc is the size adjustment factor. To apply for user job level, To automatically approve the recall volume, For the completeness of the user application form, This refers to the past approval rate of applicants.
[0058] In some optional embodiments, the automatic execution coefficients can also be calculated using the following method. :
[0059] Here, sign(x) represents the sign of x, which is 1 when x is greater than or equal to 0, and 0 otherwise.
[0060] In some optional embodiments, the automatic execution coefficient can also be used. Resources of each service node in the server cluster The authorization and approval capabilities of each service node in the computing server cluster.
[0061] For any i-th service node in the server cluster Its corresponding resources This service node Corresponding authorization and approval capabilities The calculation method is as follows:
[0062] in, Indicates the current time. Indicates the time when the process was initiated. Indicates server At the time of collection The comprehensive performance index is calculated using the following logic:
[0063] Among them, service nodes Corresponding resources ,in, For this server, time t and server startup time are... Total number of similar tasks processed internally. Let be the CPU utilization of the server at time t. This sets a switch for acquiring external resources for the server; 1 indicates on and 0 indicates off. Let be the network bandwidth utilization of the server at time t. This represents the maximum network bandwidth of the server at time t.
[0064] At this point, the server node's approval and authorization capabilities...
[0065] In some optional embodiments, the authorization approval reliability coefficient can also be used. Completeness coefficient Authorization and approval capabilities of service nodes Determine the mandatory review points for copyright approval. Actual approval nodes And complete the approval process. Traverse the list of mandatory nodes for copyright approval For any copyright approval process, a mandatory review step is required. Its actual approval process The determination method is determined through the following process: In some alternative embodiments, when the reliability coefficient Less than the minimum reliability coefficient or completeness coefficient Less than the minimum completeness coefficient Or the authorization and approval capabilities of service nodes All less than the minimum authorized approval capacity When setting parameters, manual approval is required.
[0066] In some optional embodiments, the authorization approval capability at the service node In the middle, find the reliability coefficient. Style leans towards type Infringement level Available approval nodes .
[0067] In some optional embodiments, the authorization and approval capabilities of the service nodes are traversed. For any i-th approval authorization capability ,when Add it to the available approval node when the following conditions are met. In the middle (initialized to empty by default), after traversal, available approval nodes are obtained. .
[0068] The specific conditions are: the service node's authorization and approval capabilities. The overall approval capacity of the i-th approval node is greater than or equal to That is, the following formula holds true:
[0069] The comprehensive approval capability of the i-th approval node The calculation logic is as follows:
[0070] In some optional embodiments, the available approval nodes are calculated. Joint approval capability Does it exceed the reliability coefficient? Style leans towards type Infringement level Comprehensive approval capacity requirements .
[0071] In some alternative embodiments, the system calculates the overall approval capability requirements. Assess the system's current approval capabilities to provide a basis for evaluating automated approvals, including its comprehensive approval capability requirements. The calculation logic is as follows:
[0072] in, For the number of approval nodes, This represents the factor that increases the complexity of node collaboration. This represents the comprehensive approval capacity of the k-th approval node.
[0073]
[0074] In some alternative embodiments, when At that time, by available approval nodes Implement fully automated approval.
[0075] In some alternative embodiments, when At that time, by available approval nodes The process is fully automated, and once the node approval result is "passed," it proceeds to the next approval node.
[0076] In some alternative embodiments, when At that time, by available approval nodes Perform semi-automated approval and calculate the completion rate of each node in the approval process. Filter out those below the automatic approval threshold The remaining mandatory approval nodes will be manually approved, while the rest will be approved automatically.
[0077] In some optional embodiments, for any i-th approval node The completion rate of its corresponding approval process nodes The calculation logic is as follows:
[0078] in, , , , , These are the weighting coefficients, Based on the basic completion level, its calculation logic is as follows:
[0079] in, This represents the total number of approval nodes. This represents the total number of remaining nodes. Adjust the node importance coefficient.
[0080] The calculation logic for the impact of node time efficiency completion is as follows:
[0081] in, This represents the average approval time for this node. This represents the average approval time for the process.
[0082] The calculation logic for the impact of style jump completion is as follows:
[0083] in, The node jump index, The approval style of the node is biased towards the index. Human factors are the influencing factors.
[0084] The calculation logic for the impact of the approval personnel's completion level is as follows:
[0085] in, For the level of the approver, Adjustment coefficient for the level of personnel approving nodes.
[0086] When there are i approval nodes Node completion rate Less than the automatic approval threshold At that time, the node is manually approved, and the node completion rate is... Greater than or equal to the automatic approval threshold At that time, the node is automatically approved by the system.
[0087] This disclosure extracts the core content of copyright protection documents and compresses only the core content into high-density key points. A segmented, iterative comparison algorithm based on this high-density key point compression ensures efficient retrieval while avoiding the problems of low search volume and inaccurate matching results caused by including non-copyright protection core content in the search scope. Furthermore, based on this high-density key point compression segmented, iterative comparison algorithm, while ensuring approval quality, reduces the workload of manual processing during the copyright inspection and approval process by automatically completing the approval of some nodes. Based on the same inventive concept, and corresponding to any of the above-described embodiments, this application also provides a dedicated line service data transmission device.
[0088] Figure 4 This is a schematic diagram of the structure of a copyright protection system provided in an embodiment of this application.
[0089] refer to Figure 4 As shown, the copyright protection system includes: S202: Protection point determination module, configured to: determine the core protection points of the target file to be reviewed; In some optional embodiments, module S202 further performs the following: Determine the header file and file extension of the target file; Based on the header file and the file extension, the file type of the target file is confirmed; Based on the file type, the core protection points of the target file are determined through a preset neural network model; The preset neural network model includes a format conversion layer, a feature extraction layer, and a validity filtering layer; based on the file type, the core protection points of the target file are determined through a preset multi-channel neural network model, including: Based on the file type, the target file is converted using the format conversion layer to obtain the converted target file; Based on the feature extraction layer, the features of the converted target file are extracted to obtain the core features; Based on the validity filtering layer, the core features are filtered through validity activation parameters to obtain core protection points.
[0090] The file type includes at least one of text files, image files, video files, and audio files; according to the file type, the target file is format-converted through the format conversion layer to obtain the converted target file, including: In response to determining that the file type is a text file, the text content of the target file is read and concatenated into a string, and the string is saved as a txt file; In response to determining that the file type is an image file, a preset transcoding tool is invoked to convert the target file into JPEG format; In response to determining that the file type is a video file, a preset transcoding tool is invoked to convert the target file into mp4 format; In response to determining that the file type is an audio file, a preset transcoding tool is invoked to convert the target file into mp3 format.
[0091] The feature extraction layer extracts features from the converted target file to obtain core features, including: Construct a multi-channel neural network; Determine the location encoding of the converted target file; The converted target file and the location encoding are input into the multi-channel neural network, and the core features of the converted target file and the effective location of the core features are output through the multi-channel neural network.
[0092] The constructed multi-channel neural network includes: A first multi-channel neural network is constructed using an attention mechanism; wherein the first multi-channel neural network takes the transformed target file and the position encoding as input, and outputs contextual semantic features after residual connection and attention weighting. A second multi-channel neural network is constructed using residual fully connected layers; wherein, the second multi-channel neural network is configured with the above and below layers as described above. Specifically, based on the validity filtering layer, the core features are filtered using validity activation parameters to obtain core protection points, including: Input the core features and the valid locations into the validity filtering layer; In response to the determination that the confidence level of the effective location is higher than the effectiveness activation parameter, the core feature corresponding to the effective location is determined as the core protection point.
[0093] After determining the core protection points of the target document to be reviewed, the method further includes: The core protection point is segmented according to a preset length to obtain several segmented core protection points; Fuzzy compression is performed on several of the segmented core protection points to obtain fuzzy compressed core protection points; The core protection points after fuzzy compression are cyclically calculated to obtain high-density key point compression; The high-density key point compression is subjected to product iteration processing to obtain the high-density key point compression of the target file.
[0094] Specifically, fuzzy compression is performed on several of the segmented core protection points to obtain fuzzy compressed core protection points, including: Each segmented core protection point is fuzzed to obtain several fuzzed core protection points. Local compression is performed on several of the aforementioned fuzzy core protection points to obtain several compressed core protection points; The compressed core protection points are sorted to obtain the fuzzy compressed core protection points.
[0095] Specifically, several of the blurred core protection points are locally compressed to obtain several compressed core protection points, including: Determine the average, maximum, minimum, and variance of several core protection points after fuzzing; Based on the average, maximum, minimum and variance, several core protection points after fuzzing are locally compressed to obtain several compressed core protection points.
[0096] The process involves iteratively calculating the core protection points after fuzzy compression to obtain high-density key point compression, including: The core protection points after fuzzy compression are divided into several segments according to a first preset number of bits; For each segment, take a sub-block with a second preset number of bits in sequence, and perform an XOR modulo operation on each sub-block to obtain the compressed code of each sub-block; The compressed code is concatenated to obtain high-density key point compression for each segment; The high-density key points of each segment are compressed to obtain the high-density key points of the core protection point.
[0097] The high-density key point compression is expressed as follows: ; in, This indicates high-density compression. Indicates taking The high-density compression of key points has the following calculation logic: ; in, For XOR operation, To obtain Subscript arrive The binary encoding of bmod(x,y) represents the binary encoding of x,y. press Perform the mold taking operation. This represents the modulo-mixing parameter.
[0098] The step of performing product iteration processing on the high-density key point compression to obtain the high-density key point compression of the target file includes: The high-density key point compression is multiplied and iteratively merged to obtain the merged high-density key point compression. The merged high-density key point compression is spliced to obtain the high-density key point compression of the target file.
[0099] The method further includes, after obtaining the high-density key point compression of the target file, the following: High-density key point compression is determined for each file in the target database, and a feature database is constructed based on the high-density key point compression of each file.
[0100] S204: The file list determination module is configured to: determine a list of target files whose overlap ratio with the content of the core protection point is greater than a threshold in a preset feature database; In some optional embodiments, module S204 further performs the following: The high-density key points of the target file are compressed and segmented according to a preset bit length to form a target segment set; The high-density key points of each file in the feature database are compressed and segmented according to the preset bit length to form a candidate segment set; Traverse the target segment set, compare the content similarity between the content of each segment in the target segment set and the content of each segment in the candidate segment set, and obtain the number of segments with the same content; The content overlap ratio is determined based on the number of segments and the number of segments in the target segment set and the candidate segment set; In response to determining that the content overlap ratio is greater than a threshold, a target file list is generated based on the target file.
[0101] The determination of the content overlap ratio based on the number of segments and the number of segments in the target segment set and the candidate segment set includes: Determine the number of the first segment in the target segment set and the number of the second segment in the candidate segment set; In response to determining that the number of the first segment is less than the number of the second segment, the ratio of the number of segments to the number of the first segment is used as the content overlap ratio; In response to determining that the number of the second segment is less than the number of the first segment, the ratio of the number of segments to the number of the second segment is used as the content overlap ratio.
[0102] S206: The first review node determination module is configured to determine the first review node of the target file based on the target file list.
[0103] In some optional embodiments, module S206 further performs the following: Based on the number of historical automatic approval nodes, the number of historical approval nodes, and the number of nodes that were returned after approval for each file in the target file list, the jump threshold of the approval node is determined. The jump threshold is compared with the jump index of each audit node in the preset audit node list, and the comparison result is obtained. The first audit node is determined based on the comparison results.
[0104] The determination of the first review node based on the comparison results includes: In response to determining that the jump index of any of the audit nodes is greater than or equal to the jump threshold, the audit node is designated as the first audit node.
[0105] S208: Copyright review module, configured to: determine a second review node in the first review node based on the style bias type and infringement level of the target file, and perform copyright review on the target file through the second review node.
[0106] In some optional embodiments, module S208 further performs the following: Based on the target file and its description file, determine the style bias type of the target file; The authorization approval reliability coefficient of the first review node is determined based on the style preference type; The completeness coefficient of the first review node is determined based on the infringement level. Calculate the automatic execution coefficient of the first review node based on the authorization approval reliability coefficient and the completeness coefficient; Based on the automatic execution coefficient and the resources of each service node in the server cluster, the authorization and approval capabilities of each service node are calculated. Based on the authorization approval reliability coefficient, the completeness coefficient, and the authorization approval capability, a second review node is determined in the first review node.
[0107] The copyright review of the target file through the second review node includes: For each review node in the second review node, calculate the completion degree of each review node; wherein, the node completion degree is determined by a weighted average of the basic completion degree, time efficiency completion degree, style jump completion degree, and approver level completion degree. The target file is reviewed for copyright based on the completion level, and the final review result is obtained.
[0108] The process includes copyright verification of the target file based on the completion level, and obtaining the final verification result, including: In response to determining that the completion level is greater than or equal to a preset approval threshold, the target file is subjected to copyright review based on the target server, and a first review result is obtained; In response to determining that the completion rate is less than the approval threshold, the target file is manually reviewed for copyright, and a second copyright review result is obtained; Based on the first review result and the second review result, the final review result of the target file is generated.
[0109] For ease of description, the above devices are described in terms of function, divided into various modules. Of course, in implementing this application, the functions of each module can be implemented in one or more software and / or hardware.
[0110] The apparatus described above is used to implement the corresponding leased line service data transmission method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0111] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for transmitting leased line service data as described in any of the above embodiments.
[0112] Figure 5 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 810, a memory 820, an input / output interface 830, a communication interface 840, and a bus 850. The processor 810, memory 820, input / output interface 830, and communication interface 840 are interconnected internally via the bus 850. The processor 810 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.
[0113] The memory 820 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 820 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 820 and is called and executed by the processor 810.
[0114] The input / output interface 830 is used to connect input / output modules to enable information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.
[0115] The communication interface 840 is used to connect the communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 810, memory 820, input / output interface 830, and communication interface 840.
[0116] It should be noted that although the above-described device only shows the processor 810, memory 820, input / output interface 830, communication interface 840, and bus 850, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures. The electronic devices described above are used to implement the corresponding leased line service data transmission method in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here. Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the leased line service data transmission method as described in any of the above embodiments.
[0117] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device. The aforementioned non-transitory computer-readable storage media can be any available medium or data storage device that a computer can access, including but not limited to magnetic storage (e.g., floppy disks, hard disks, magnetic tapes, magneto-optical disks (MOs), etc.), optical storage (e.g., CDs, DVDs, BDs, HVDs, etc.), and semiconductor storage (e.g., ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs)). The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the leased line service data transmission method as described in any of the embodiments in the exemplary method section above, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0118] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this application (including the claims) is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in the details for the sake of brevity.
[0119] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this application, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this application, and this also takes into account the fact that the details of the implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this application will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this application, it will be apparent to those skilled in the art that the embodiments of this application can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.
[0120] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.
[0121] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.
Claims
1. A copyright review method, characterized in that, include: Identify the core protection points of the target documents to be reviewed; In a preset feature database, a list of target files whose content overlaps with the core protection point content by a proportion greater than a threshold is determined. The first review node for the target file is determined based on the target file list; Based on the style preference type and infringement level of the target file, a second review node is determined in the first review node, and the copyright review of the target file is carried out through the second review node.
2. The method according to claim 1, characterized in that, The process of determining the core protection points of the target document to be reviewed includes: Determine the header file and file extension of the target file; Based on the header file and the file extension, the file type of the target file is confirmed; Based on the file type, the core protection points of the target file are determined through a preset neural network model.
3. The method according to claim 2, characterized in that, The preset neural network model includes a format conversion layer, a feature extraction layer, and a validity filtering layer; The step of determining the core protection points of the target file based on the file type using a preset multi-channel neural network model includes: Based on the file type, the target file is converted using the format conversion layer to obtain the converted target file; Based on the feature extraction layer, the features of the converted target file are extracted to obtain the core features; Based on the validity filtering layer, the core features are filtered through validity activation parameters to obtain core protection points.
4. The method according to claim 3, characterized in that, The file types include at least one of text files, image files, video files, and audio files; The step of converting the target file according to the file type through the format conversion layer to obtain the converted target file includes: In response to determining that the file type is a text file, the text content of the target file is read and concatenated into a string, and the string is saved as a txt file; In response to determining that the file type is an image file, a preset transcoding tool is invoked to convert the target file into JPEG format; In response to determining that the file type is a video file, a preset transcoding tool is invoked to convert the target file into mp4 format; In response to determining that the file type is an audio file, a preset transcoding tool is invoked to convert the target file into mp3 format.
5. The method according to claim 3, characterized in that, The feature extraction layer extracts features from the converted target file to obtain core features, including: Construct a multi-channel neural network; Determine the position encoding of the converted target file; The converted target file and the location encoding are input into the multi-channel neural network, and the core features of the converted target file and the effective location of the core features are output through the multi-channel neural network.
6. The method according to claim 5, characterized in that, The construction of the multi-channel neural network includes: A first multi-channel neural network is constructed using an attention mechanism; wherein the first multi-channel neural network takes the transformed target file and the position encoding as input, and outputs contextual semantic features after residual connection and attention weighting. A second multi-channel neural network is constructed using residual fully connected layers; wherein the second multi-channel neural network takes the contextual semantic features as input, is located by the residual fully connected layers, and outputs the core features and the effective position of the core features.
7. The method according to claim 5, characterized in that, The core protection points are obtained by filtering the core features based on the validity filtering layer using validity activation parameters, including: Input the core features and the valid locations into the validity filtering layer; In response to the determination that the confidence level of the effective location is higher than the effectiveness activation parameter, the core feature corresponding to the effective location is determined as the core protection point.
8. The method according to claim 1, characterized in that, After determining the core protection points of the target file to be reviewed, the method further includes: The core protection point is segmented according to a preset length to obtain several segmented core protection points; Fuzzy compression is performed on several of the segmented core protection points to obtain fuzzy compressed core protection points; The core protection points after fuzzy compression are cyclically calculated to obtain high-density key point compression; The high-density key point compression is subjected to product iteration processing to obtain the high-density key point compression of the target file.
9. The method according to claim 8, characterized in that, The step of performing fuzzy compression on several segmented core protection points to obtain fuzzy compressed core protection points includes: Each segmented core protection point is fuzzed to obtain several fuzzed core protection points. Local compression is performed on several of the aforementioned fuzzy core protection points to obtain several compressed core protection points; The compressed core protection points are sorted to obtain the fuzzy compressed core protection points.
10. The method according to claim 9, characterized in that, The process of locally compressing the several blurred core protection points yields several compressed core protection points, including: Determine the average, maximum, minimum, and variance of several core protection points after fuzzing; Based on the average, maximum, minimum and variance, several core protection points after fuzzing are locally compressed to obtain several compressed core protection points.
11. The method according to claim 8, characterized in that, The step of iteratively calculating the core protection points after fuzzy compression to obtain high-density key point compression includes: The core protection points after fuzzy compression are divided into several segments according to a first preset number of bits; For each segment, take a sub-block with a second preset number of bits in sequence, and perform an XOR modulo operation on each sub-block to obtain the compressed code of each sub-block; The compressed code is concatenated to obtain high-density key point compression for each segment; The high-density key points of each segment are compressed to obtain the high-density key points of the core protection point.
12. The method according to claim 11, characterized in that, The high-density key point compression is expressed as follows: ； in, This indicates high-density compression. Indicates taking The high-density compression of key points has the following calculation logic: ； in, For XOR operation, To obtain Subscript arrive The binary encoding of bmod(x,y) represents the binary encoding of x,y. press Perform the mold taking operation. This represents the modulo-mixing parameter.
13. The method according to claim 12, characterized in that, The step of performing product iteration processing on the high-density key point compression to obtain the high-density key point compression of the target file includes: The high-density key point compression is multiplied and iteratively merged to obtain the merged high-density key point compression. The merged high-density key point compression is spliced to obtain the high-density key point compression of the target file.
14. The method according to claim 8, characterized in that, After obtaining the high-density key point compression of the target file, the method further includes: High-density key point compression is determined for each file in the target database, and a feature database is constructed based on the high-density key point compression of each file.
15. The method according to claim 14, characterized in that, The step of determining a list of target files in a preset feature database whose content overlaps with the core protection point by a proportion greater than a threshold includes: The high-density key points of the target file are compressed and segmented according to a preset bit length to form a target segment set; The high-density key points of each file in the feature database are compressed and segmented according to the preset bit length to form a candidate segment set; Traverse the target segment set, compare the content similarity between the content of each segment in the target segment set and the content of each segment in the candidate segment set, and obtain the number of segments with the same content; The content overlap ratio is determined based on the number of segments and the number of segments in the target segment set and the candidate segment set; In response to determining that the content overlap ratio is greater than a threshold, a target file list is generated based on the target file.
16. The method according to claim 15, characterized in that, The step of determining the content overlap ratio based on the number of segments and the number of segments in the target segment set and the candidate segment set includes: Determine the number of the first segment in the target segment set and the number of the second segment in the candidate segment set; In response to determining that the number of the first segment is less than the number of the second segment, the ratio of the number of segments to the number of the first segment is used as the content overlap ratio; In response to determining that the number of the second segment is less than the number of the first segment, the ratio of the number of segments to the number of the second segment is used as the content overlap ratio.
17. The method according to claim 1, characterized in that, The step of determining the first review node of the target file based on the target file list includes: Based on the number of historical automatic approval nodes, the number of historical approval nodes, and the number of nodes that were returned after approval for each file in the target file list, the jump threshold of the approval node is determined. The jump threshold is compared with the jump index of each audit node in the preset audit node list, and the comparison result is obtained. The first audit node is determined based on the comparison results.
18. The method according to claim 17, characterized in that, The process of determining the first audit node based on the comparison results includes: In response to determining that the jump index of any of the audit nodes is greater than or equal to the jump threshold, the audit node is designated as the first audit node.
19. The method according to claim 18, characterized in that, The process of determining the second review node in the first review node based on the style bias type and infringement level of the target file includes: Based on the target file and its description file, determine the style bias type of the target file; The authorization approval reliability coefficient of the first review node is determined based on the style preference type; The completeness coefficient of the first review node is determined based on the infringement level. Calculate the automatic execution coefficient of the first review node based on the authorization approval reliability coefficient and the completeness coefficient; Based on the automatic execution coefficient and the resources of each service node in the server cluster, the authorization and approval capabilities of each service node are calculated. Based on the authorization approval reliability coefficient, the completeness coefficient, and the authorization approval capability, a second review node is determined in the first review node.
20. The method according to claim 19, characterized in that, The copyright review of the target file through the second review node includes: For each review node in the second review node, calculate the completion degree of each review node; wherein, the node completion degree is determined by a weighted average of the basic completion degree, time efficiency completion degree, style jump completion degree, and approver level completion degree. The target file is reviewed for copyright based on the completion level, and the final review result is obtained.
21. The method according to claim 20, characterized in that, The copyright review of the target file based on the completion level, and the obtaining of the final review result, includes: In response to determining that the completion level is greater than or equal to a preset approval threshold, the target file is subjected to copyright review based on the target server, and a first review result is obtained; In response to determining that the completion rate is less than the approval threshold, the target file is manually reviewed for copyright, and a second copyright review result is obtained; Based on the first review result and the second review result, the final review result of the target file is generated.
22. A copyright verification system, characterized in that, include: The protection point determination module is configured to: determine the core protection points of the target file to be reviewed; The file list determination module is configured to: determine a list of target files whose overlap ratio with the content of the core protection point is greater than a threshold in a preset feature database; The first review node determination module is configured to: determine the first review node of the target file based on the target file list; The copyright review module is configured to: determine a second review node in the first review node based on the style preference type and infringement level of the target file, and conduct copyright review on the target file through the second review node.
23. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the program to implement the method of claims 1-21.
24. A computer-readable storage medium, characterized in that, The medium stores computer instructions for causing the computer to perform the method described in claims 1-21.
25. A computer program product, characterized in that, It includes one or more computer programs that, when executed by one or more processors, implement the method as described in any one of claims 1-21.