A method for screening broadcast content for content censorship

Abstract

The present invention provides a method for screening broadcast content for content review, and relates to the technical field of broadcast content review. The method includes collecting historical broadcast content review data, establishing a broadcast content historical review database; extracting historical illegal broadcast content and historical normal broadcast content of different content types in the broadcast content historical review database, performing feature extraction, and forming content type broadcast review feature information; extracting historical missed review broadcast content of different content types in the broadcast content historical review database, performing feature information extraction, and forming missed review comment feature data; acquiring real-time broadcast content, and performing review screening to form reviewed and screened real-time broadcast content; collecting review and screened comment information of real-time broadcast content, and performing supplementary screening based on missed review comment feature data to form complete screening of real-time broadcast content. The method uses big data to perform more efficient and accurate, sufficient and reasonable content screening.

Description

Technical Field

[0001] The present invention relates to the technical field of broadcast content review, and in particular to a broadcast content screening method for content review. Background Art

[0002] Broadcast content is usually screened before being broadcast. One of the important tasks of screening is to conduct a reasonable review of the content to identify content that violates regulations and avoid negative impacts on the social level after broadcasting.

[0003] Usually, content screening is mostly done manually, which is inefficient and cannot guarantee high accuracy. Of course, there is also artificial intelligence screening, but the current artificial intelligence mainly collects a large amount of data for feature extraction. Each screening analysis requires the use of a large amount of feature information data, resulting in a waste of resources and reduced efficiency. Of course, artificial intelligence also has certain aspects of unreasonable algorithms that lead to missed reviews. This kind of missed review can only be screened manually by releasing first and then reviewing, which increases the screening cost and has a negative social impact.

[0004] Therefore, designing a broadcast content screening method for content review, which can make use of big data to conduct more efficient, accurate, sufficient and reasonable content screening, is an urgent problem to be solved. Summary of the invention

[0005] The purpose of the present invention is to provide a method for screening broadcast content for content review, by collecting historical review data of broadcast content, establishing a database for feature extraction, and extracting feature information based on different types of broadcast content. Compared with directly extracting feature information, this can greatly reduce the scope of feature information collection. On the one hand, it can improve the accuracy of feature information extraction, which is conducive to improving the accuracy of subsequent content screening. On the other hand, because it is not performed by type, it also greatly reduces the efficiency of subsequent screening and analysis. At the same time, in view of the missed reviews caused by the imperfection of intelligent algorithms, the missed review content screening feature information based on comment information is established, which is conducive to the subsequent use as a supplement to intelligent screening, greatly improving the efficiency and accuracy of broadcast content screening, and fully ensuring the rationality of content screening.

[0006] In the first aspect, the present invention provides a method for screening broadcast content for content review, including collecting historical broadcast content review data and establishing a broadcast content historical review database; extracting historical illegal broadcast content and historical normal broadcast content of different content types in the broadcast content historical review database, performing feature extraction based on content type, and forming content type broadcast review feature information; extracting historical missed broadcast content of different content types in the broadcast content historical review database, performing feature information extraction based on comments, and forming missed review comment feature data; acquiring real-time broadcast content, and performing review screening according to the content type broadcast review feature information, and forming reviewed and screened real-time broadcast content; collecting review and screened comment information of real-time broadcast content, and performing supplementary screening according to the missed review comment feature data, and forming a complete screening of real-time broadcast content.

[0007] In the present invention, the method collects historical review data of broadcast content, establishes a database for feature extraction, and extracts feature information based on different types of broadcast content. Compared with directly extracting feature information, this can greatly reduce the scope of feature information collection. On the one hand, it can improve the accuracy of feature information extraction, which is conducive to improving the accuracy of subsequent content screening. On the other hand, because it is not performed by type, it also greatly reduces the efficiency of subsequent screening and analysis. At the same time, in view of the missed reviews caused by the imperfection of intelligent algorithms, the missed review content screening feature information based on comment information is established, which is conducive to the subsequent use as a supplement to intelligent screening, greatly improving the efficiency and accuracy of broadcast content screening, and fully ensuring the rationality of content screening.

[0008] As a possible implementation method, historical illegal broadcast content and historical normal broadcast content of different content types in the broadcast content historical review database are extracted, and feature extraction based on content type is performed to form content type broadcast review feature information, including: extracting historical illegal broadcast content and historical normal broadcast content of the same content type in the broadcast content historical review database to form type feature basic content data corresponding to the content type; extracting illegal feature information of all historical illegal broadcast content in the type feature basic content data to form type illegal content review feature information; based on the type illegal content review feature information, extracting supplementary normal feature information of all historical normal broadcast content in the type feature basic content data to form type supplementary normal content review feature information; and combining type illegal content review feature information and type supplementary normal content review feature information to form content type broadcast review feature information.

[0009] In the present invention, the feature information of the broadcast content is directly extracted for content screening, which will reduce the efficiency of the analysis because the amount of feature information extracted is too large each time the feature information is extracted for analysis and judgment. At the same time, the feature information in advance will not have better directionality because there is no targeted basic data. Therefore, the feature information of the collected historical review data is extracted, and the content type needs to be divided first. Regarding the content type, it can be a broadcast category, such as news, literature, education, etc., or it can be a category for the industry, such as sports, manufacturing, finance, etc. The feature information extraction based on the industry division is more directional for subsequent comments. In addition, for the feature information extraction, it is necessary to consider the feature information of the violation judgment and the situation where there is violation information but it is regarded as normal content. After all, some sensitive words will be transformed into other meanings when they are raised in the form of rhetorical questions. For different information such as images and sounds, there are also situations where there are sensitive words but they are regarded as normal content. Therefore, after extracting the feature information of the illegal content, the present application also extracts features for normal broadcasts with sensitive content to supplement the deviation of the illegal feature judgment.

[0010] As a possible implementation method, violation feature information of all historical illegal broadcast contents in the type feature basic content data is extracted to form type illegal content review feature information, including: extracting violation information of different historical illegal broadcast contents in the type feature basic content data to form type initial violation information; comparing and clustering different types of initial violation information in the following manner: if there is identical information between different types of initial violation information, extracting the identical part of the information as clustering basic information, obtaining the remaining necessary information in the different types of initial violation information except the clustering basic information to form a remaining necessary information set, and combining the clustering basic information and the corresponding remaining necessary information set to form clustering violation feature information; and aggregating different clustering violation feature information to form type illegal content review feature information corresponding to the content type.

[0011] In the present invention, the feature information extraction of illegal broadcasting content is mainly formed based on the illegal information marked under the historical review content. It is worth noting that for illegal information, there are different changes of a certain type of illegal information under big data. Therefore, when extracting, the most recent feature information is determined by using the same information, and then the necessary combined feature information is extracted for different changes, thereby forming different necessary combined feature data under the basic feature information to ensure the integrity and sufficiency of the same type of illegal feature information. In addition, using the form of clustering basic information and the remaining necessary information set as feature information can greatly reduce the storage space of feature information, which is conducive to saving resources.

[0012] As a possible implementation method, based on the type violation content review feature information, supplementary normal feature information is extracted for all historical normal broadcast contents in the type feature basic content data to form type supplementary normal content review feature information, including: extracting the following supplementary normal feature information for different clustered violation feature information of the type violation content review feature information: if normal information containing type initial violation information exists in the historical normal broadcast content, then the normal information with an amount of information exceeding the corresponding type initial violation information is extracted to form type supplementary normal review content under the type violation content review feature information corresponding to the type initial violation information; different types of supplementary normal review content under the group type violation content review feature information are gathered to form the corresponding type supplementary normal content review feature information.

[0013] In the present invention, it can be understood that even if the broadcast content contains partial or complete information of illegal characteristics, it does not necessarily mean that the broadcast content is illegal. For example, for text broadcasts, sensitive words and phrases with illegal characteristics may be converted into normal content in the form of rhetorical questions. For image data, partial reasonable obstruction on the illegal content may convert sensitive information, etc. Therefore, in order to avoid misjudgment of these contents by directly using illegal characteristic information for judgment, supplementing the normal review content characteristic information is to extract more information containing illegal characteristic information from the normal broadcast content, so as to form characteristic information data for accurately judging whether such broadcast content is illegal.

[0014] As a possible implementation method, historical missed-review broadcast contents of different content types in a broadcast content historical review database are extracted, and feature information extraction based on comments is performed to form missed-review comment feature data, including: for different historical missed-review broadcast contents of the same content type, corresponding comment information is extracted to perform word repetition statistics, and repetitive words are clustered to form a type repetitive word data set; based on the type repetitive word data set, repetitive word groups of missed-review comments that exist in different historical missed-review broadcast contents are determined; repetition counts are performed on different missed-review comment repetitive word groups in corresponding historical missed-review broadcast contents to determine the repetitive word quantities corresponding to different repetitive words in the missed-review comment repetitive word groups; feature information of missed-review comments is extracted based on different missed-review comment repetitive word groups and the repetitive word quantities of different repetitive words corresponding to the missed-review comment repetitive word groups to form missed-review comment feature data.

[0015] In the present invention, although the review of broadcast content is carried out in an intelligent form, which greatly improves the efficiency and accuracy of screening, there are always loopholes in the algorithm, resulting in the omission of some broadcast content with relatively vague features. And these missed broadcast content will cause social effects because of these illegal contents. The most direct expression of this social effect is the comment information on the broadcast content. Therefore, in order to supplement the algorithm loopholes for content screening and ensure the accuracy and rationality of screening, the content screened by feature information is subjected to historical review and the part determined to be illegal is subjected to feature information extraction for the comment information. The features extracted here mainly consider highly repetitive words and the combination of these sensitive words. After all, unrelated broadcast content must have the same features in the comment information, especially in these illegal contents, and highly repetitive information must be the comment feature information formed by the negative impact. Of course, there is still a process of quantitative change to qualitative change in judging based on the comment information, so it is necessary to count the combination of repetitive words and the amount of repetition to accurately grasp the feature data.

[0016] As a possible implementation method, characteristic information of missed comments is extracted according to different repetitive vocabulary groups of missed comments and the repetitive vocabulary quantities of different repetitive vocabulary corresponding to the repetitive vocabulary groups of missed comments to form characteristic data of missed comments, including: if different repetitive vocabulary groups of missed comments have an inclusion relationship, then the repetitive vocabulary group of missed comments with the least number of repetitive vocabulary is determined as the characteristic group of missed comments, and the maximum repetitive quantity corresponding to each repetitive vocabulary is determined as the characteristic quantity of the vocabulary of missed comments; if the repetitive vocabulary group of missed comments does not have an inclusion relationship with other repetitive vocabulary groups of missed comments, then the repetitive vocabulary group of missed comments is determined as the characteristic group of missed comments, and the repetitive quantity of each corresponding repetitive vocabulary is determined as the characteristic quantity of the vocabulary of missed comments; different characteristic groups of missed comments and the characteristic quantities of missed comments corresponding to different repetitive vocabulary in the characteristic groups of missed comments are gathered to form characteristic data of missed comments.

[0017] In the present invention, of course, the features extracted from the comment information of the illegal content that was missed under big data also have the sameness and inclusiveness, so after extracting the feature information of different missed content, reasonable merging and optimization can be performed. It can be understood that for the adjustment information with an inclusive relationship, the feature information with the smallest range is the minimum judgment standard for determining whether the missed content is illegal, and the maximum amount of repetition is extracted as the quantitative judgment standard because if the missed content is really illegal, the repetition amount of these feature words will inevitably reach as the amount of comment information increases, and selecting a smaller value may cause misjudgment.

[0018] As a possible implementation method, real-time broadcast content is obtained, and review and screening is performed according to the content type broadcast review feature information to form review and screening real-time broadcast content, including: obtaining real-time broadcast content, extracting the corresponding type of illegal content review feature information, and performing review and screening in the following manner: if there is no combination of clustering basic information and remaining necessary information under any clustering illegal feature information in the real-time broadcast content, then the corresponding real-time broadcast content is retained and marked as review and screening real-time broadcast content; if there is any combination of clustering basic information and remaining necessary information under any clustering illegal feature information in the real-time broadcast content, then the corresponding type of supplementary normal review content is combined for review and screening analysis to form a supplementary review and screening result.

[0019] In the present invention, after extracting the characteristic information for content screening, the real-time broadcast content can be screened and judged. The first choice is to judge based on the extracted illegal characteristic information. It can be understood that as long as the real-time broadcast content does not involve illegal characteristic information, it can be directly determined to be normal broadcast content. Broadcast content with illegal characteristic information needs to be combined with supplementary normal review content for special situation analysis and judgment.

[0020] As a possible implementation method, if there is a combination of clustering basic information and remaining necessary information under any clustering violation feature information in the real-time broadcast content, the review screening analysis is performed in combination with the corresponding type-supplemented normal review content to form a supplementary review screening result, including: determining the combination of clustering basic information and remaining necessary information existing in the real-time broadcast content, and traversing the type-supplemented normal review content: if there is normal information in the real-time broadcast content that is the same as that in the type-supplemented normal review content, and it contains a combination of clustering basic information and remaining necessary information, then the corresponding real-time broadcast content is retained and marked as review-screened real-time broadcast content; if there is no normal information in the real-time broadcast content that is the same as that in the type-supplemented normal review content, then the corresponding real-time broadcast content is deleted.

[0021] In the present invention, when using the supplementary normal review content to judge the real-time broadcast content with illegal characteristic information, it is mainly to determine whether the expanded information amount of the illegal information in the real-time broadcast content exists in the supplementary normal review content. If it exists, it can be determined that although the real-time broadcast content contains illegal characteristics, it does not touch the requirements of violation as a whole.

[0022] As a possible implementation method, the comment information of the real-time broadcast content is collected, reviewed and screened, and supplementary screening is performed based on the characteristic data of the missed comments to form a complete screen of the real-time broadcast content, including: collecting, reviewing and screening the comment information of the real-time broadcast content, and decomposing the comment information into words to form a decomposed vocabulary set; based on the decomposed vocabulary set, supplementary screening and analysis is performed in combination with the characteristic data of the missed comments of the corresponding content type to determine the complete screen of the real-time broadcast content.

[0023] In the present invention, the real-time broadcast content that has been screened for illegal characteristic information needs to be reasonably monitored to avoid missing illegal content due to algorithm defects. The screening mainly involves continuously acquiring comment information and then breaking down the vocabulary to compare and analyze the characteristics of the missed comments.

[0024] As a possible implementation method, supplementary screening and analysis is performed based on the disassembled vocabulary set and combined with the characteristic data of missed comments of the corresponding content type to determine the complete screening of real-time broadcast content, including: if the review information of the real-time broadcast content that is screened contains all repetitive words of any repetitive vocabulary group of missed comments, and the repetition amount of all repetitive words does not meet the requirement that the repetition amount of all repetitive words reaches the corresponding repetitive vocabulary amount, it is marked as primary missed broadcast content; if the review information of the real-time broadcast content that is screened does not contain all repetitive words of any repetitive vocabulary group of missed comments, but some of the repetitive words it has meet the requirement that the repetition amount of all repetitive words reaches the corresponding repetitive vocabulary amount, it is marked as important missed broadcast content; if the review information of the real-time broadcast content that is screened does not contain all repetitive words of any repetitive vocabulary group of missed comments, and some of the repetitive words that exist do not meet the requirement that the repetition amount of all repetitive words reaches the corresponding repetitive vocabulary amount, it is marked as secondary missed broadcast content; if the review information of the real-time broadcast content that is screened does not contain any repetitive words of any repetitive vocabulary group of missed comments, it is marked as normal real-time broadcast content.

[0025] In the present invention, since the feature analysis of missed comments is a continuous process, the analysis is to calibrate and continuously observe different comparison situations. Of course, for the comment information that has never had any missed comment feature information, it can be completely determined that the corresponding broadcast real-time content is normal. For the comment data with missed comment feature information, different feature occurrence situations are considered for classification and calibration, so as to provide key and secondary operation distinctions for subsequent continuous observation, so as to achieve reasonable use of resources.

[0026] The beneficial effects of the broadcast content screening method for content review provided by the present invention are as follows:

[0027] This method collects historical review data of broadcast content, establishes a database for feature extraction, and extracts feature information based on different types of broadcast content. Compared with direct feature information extraction, this method can greatly reduce the scope of feature information collection. On the one hand, it can improve the accuracy of feature information extraction, which is conducive to improving the accuracy of subsequent content screening. On the other hand, because it is not based on type, it also greatly reduces the efficiency of subsequent screening and analysis. At the same time, in view of the missed reviews caused by the imperfection of intelligent algorithms, the missed review content screening feature information based on comment information is established, which is conducive to the subsequent supplement of intelligent screening, greatly improving the efficiency and accuracy of broadcast content screening, and fully ensuring the rationality of content screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments of the present invention are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0029] Figure 1 A step diagram of a method for screening broadcast content for content review provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present invention will be described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0031] Broadcast content is usually screened before being broadcast. One of the important tasks of screening is to conduct a reasonable review of the content to identify content that violates regulations and avoid negative impacts on the social level after broadcasting.

[0032] Usually, content screening is mostly done manually, which is inefficient and cannot guarantee high accuracy. Of course, there is also artificial intelligence screening, but the current artificial intelligence mainly collects a large amount of data for feature extraction. Each screening analysis requires the use of a large amount of feature information data, resulting in a waste of resources and reduced efficiency. Of course, artificial intelligence also has certain aspects of unreasonable algorithms that lead to missed reviews. This kind of missed review can only be screened manually by releasing first and then reviewing, which increases the screening cost and has a negative social impact.

[0033] refer to Figure 1The embodiment of the present invention provides a method for screening broadcast content for content review. The method collects historical review data of broadcast content, establishes a database for feature extraction, and extracts feature information based on different types of broadcast content. Compared with directly extracting feature information, this can greatly reduce the scope of feature information collection. On the one hand, it can improve the accuracy of feature information extraction, which is conducive to improving the accuracy of subsequent content screening. On the other hand, because it is not performed on a type basis, it also greatly reduces the efficiency of subsequent screening and analysis. At the same time, in order to address missed reviews caused by imperfect intelligent algorithms, feature information for missed content screening based on comment information is established, which is conducive to subsequent use as a supplement to intelligent screening, greatly improving the efficiency and accuracy of broadcast content screening, and fully ensuring the rationality of content screening.

[0034] The method for screening broadcast content for content review specifically includes the following steps:

[0035] S1: Collect historical broadcast content review data and establish a broadcast content historical review database.

[0036] The establishment of a database is conducive to the updating of big data and the reasonable and efficient extraction and processing of data.

[0037] S2: Extract historical illegal broadcast contents and historical normal broadcast contents of different content types from the broadcast content historical review database, perform feature extraction based on content types, and form content type broadcast review feature information.

[0038] Extract historical illegal broadcast content and historical normal broadcast content of different content types from the broadcast content historical review database, perform feature extraction based on content type, and form content type broadcast review feature information, including: extract historical illegal broadcast content and historical normal broadcast content of the same content type from the broadcast content historical review database to form type feature basic content data corresponding to the content type; extract illegal feature information from all historical illegal broadcast content in the type feature basic content data to form type illegal content review feature information; extract supplementary normal feature information from all historical normal broadcast content in the type feature basic content data based on the type illegal content review feature information to form type supplementary normal content review feature information; and combine type illegal content review feature information and type supplementary normal content review feature information to form content type broadcast review feature information.

[0039] Directly extracting feature information from broadcast content for content screening will reduce the efficiency of analysis because the amount of feature information extracted each time for analysis and judgment will be too large, and the feature information in advance will not have better directionality because there is no targeted basic data. Therefore, to extract feature information from the collected historical review data, it is necessary to first divide the content type. Regarding the content type, it can be a broadcast category, such as news, literature, education, etc., or it can be a category for the industry, such as sports, manufacturing, finance, etc. The industry-based division facilitates the extraction of feature information for subsequent comments and is more directional. In addition, for feature information extraction, it is necessary to consider the feature information of violation judgment and the situation where there is violation information but it is regarded as normal content. After all, some sensitive words will be transformed into other meanings when they are raised in the form of rhetorical questions. For different information such as images and sounds, there are also situations where there are sensitive words but they are regarded as normal content. Therefore, after extracting the feature information of the illegal content, this application also extracts features for normal broadcasts with sensitive content to supplement the deviation of the illegal feature judgment.

[0040] Extract violation feature information of all historical illegal broadcast contents in the type feature basic content data to form type violation content review feature information, including: extracting violation information of different historical illegal broadcast contents in the type feature basic content data to form type initial violation information; performing comparative clustering on different types of initial violation information in the following manner: if there is identical information between different types of initial violation information, extract the identical part of the information as clustering basic information, obtain the remaining necessary information in the different types of initial violation information except the clustering basic information to form a remaining necessary information set, and combine the clustering basic information and the corresponding remaining necessary information set to form cluster violation feature information; aggregate different cluster violation feature information to form type violation content review feature information corresponding to the content type.

[0041] The feature information extraction of illegal broadcasting content is mainly based on the illegal information marked under the historical review content. It is worth noting that for illegal information, there are different changes of a certain type of illegal information under big data. Therefore, when extracting, the most recent feature information is determined by using the same information, and then the necessary combined feature information is extracted for different changes, and then different necessary combined feature data under the basic feature information are formed to ensure the integrity and sufficiency of the same type of illegal feature information. In addition, using the form of clustering basic information and the remaining necessary information set as feature information can greatly reduce the storage space of feature information, which is conducive to saving resources.

[0042] According to the type-violation content review feature information, supplementary normal feature information is extracted for all historical normal broadcast contents in the type-feature basic content data to form type-supplemented normal content review feature information, including: extracting the following supplementary normal feature information for different clustered violation feature information of the type-violation content review feature information: if normal information containing type-initial violation information exists in the historical normal broadcast content, then the normal information with an amount of information exceeding the corresponding type-initial violation information is extracted to form type-supplemented normal review content under the type-violation content review feature information corresponding to the type-initial violation information; different types of supplementary normal review content under the type-violation content review feature information are aggregated to form the corresponding type-supplemented normal content review feature information.

[0043] It is understandable that even if partial or complete information of illegal characteristics exists in the broadcast content, it does not necessarily mean that the broadcast content is illegal. For example, for text broadcasts, sensitive words and phrases with illegal characteristics may be converted into normal content in the form of rhetorical questions. For image data, partial reasonable obstruction of illegal content may convert sensitive information, etc. Therefore, in order to avoid misjudgment of these contents by directly using illegal characteristic information for judgment, supplementing the normal review content characteristic information is to extract more information containing illegal characteristic information from the normal broadcast content, so as to form characteristic information data that can accurately judge whether such broadcast content is illegal.

[0044] S3: extract historical missed-review broadcast contents of different content types from the broadcast content historical review database, extract feature information based on comments, and form missed-review comment feature data.

[0045] Extract historical missed-review broadcast contents of different content types from a broadcast content historical review database, perform comment-based feature information extraction, and form missed-review comment feature data, including: extracting corresponding comment information for different historical missed-review broadcast contents of the same content type, performing word repetition statistics, clustering repetitive words, and forming a type repetitive word data set; determining the missed-review comment repetitive word groups existing in different historical missed-review broadcast contents based on the type repetitive word data set; performing repetition statistics for different missed-review comment repetitive word groups under corresponding historical missed-review broadcast contents, and determining the repetitive word quantities corresponding to different repetitive words in the missed-review comment repetitive word groups; extracting missed-review comment feature information based on different missed-review comment repetitive word groups and the repetitive word quantities of different repetitive words corresponding to the missed-review comment repetitive word groups, and forming missed-review comment feature data.

[0046] Although the intelligent review of broadcast content has greatly improved the efficiency and accuracy of screening, there are always loopholes in the algorithm, which leads to the omission of some broadcast content with vague features. These missed broadcast content will cause social effects due to these illegal contents. The most direct expression of this social effect is the comment information on the broadcast content. Therefore, in order to supplement the algorithm loopholes in content screening and ensure the accuracy and rationality of screening, the content that has been screened by feature information is subjected to historical review and the feature information of the part that is determined to be illegal is extracted for the comment information. The features extracted here mainly consider highly repetitive words and the combination of these sensitive words. After all, unrelated broadcast content must have the same features in the comment information, especially in these illegal contents, highly repetitive information must be the comment feature information formed by negative impact. Of course, there is still a process of quantitative change to qualitative change in judging based on comment information, so it is necessary to count the combination of repetitive words and the amount of repetition to accurately grasp the feature data.

[0047] The characteristic information of missed comments is extracted according to different repetitive vocabulary groups of missed comments and the repetitive vocabulary quantities of different repetitive vocabulary corresponding to the repetitive vocabulary groups of missed comments to form characteristic data of missed comments, including: if different repetitive vocabulary groups of missed comments have an inclusion relationship, then the repetitive vocabulary group of missed comments with the least number of repetitive vocabulary is determined as the characteristic group of missed comments, and the maximum repetitive quantity corresponding to each repetitive vocabulary is determined as the characteristic quantity of the vocabulary of missed comments; if the repetitive vocabulary group of missed comments does not have an inclusion relationship with other repetitive vocabulary groups of missed comments, then the repetitive vocabulary group of missed comments is determined as the characteristic group of missed comments, and the repetitive quantity of each corresponding repetitive vocabulary is determined as the characteristic quantity of the vocabulary of missed comments; different characteristic groups of missed comments and the characteristic quantities of missed comments corresponding to different repetitive vocabulary in the characteristic groups of missed comments are gathered to form characteristic data of missed comments.

[0048] Of course, the features extracted from the comments of the illegal content that was missed under big data also have the same and inclusive characteristics, so after extracting the feature information of different missed content, reasonable merging and optimization can be performed. It can be understood that for the adjustment information with an inclusive relationship, the feature information with the smallest range is the minimum judgment standard for determining whether the missed content is illegal, and extracting the maximum amount of repetition as the quantitative judgment standard is to consider that if the missed content is really illegal, the repetition of these feature words will inevitably reach as the amount of comment information increases, and selecting a smaller value may cause misjudgment.

[0049] S4: Acquire real-time broadcast content, and perform review and screening based on the content type broadcast review feature information to form review and screen real-time broadcast content.

[0050] Real-time broadcast content is obtained, and review and screen it according to the content type broadcast review feature information to form review and screened real-time broadcast content, including: obtaining real-time broadcast content, extracting the corresponding type of illegal content review feature information, and conducting review and screening in the following manner: if there is no combination of clustering basic information and remaining necessary information under any clustering illegal feature information in the real-time broadcast content, then the corresponding real-time broadcast content is retained and marked as review and screened real-time broadcast content; if there is any combination of clustering basic information and remaining necessary information under any clustering illegal feature information in the real-time broadcast content, then the corresponding type of supplementary normal review content is combined for review and screening analysis to form a supplementary review and screening result.

[0051] After extracting the characteristic information for content screening, the real-time broadcast content can be screened and judged. The first choice is to judge based on the extracted illegal characteristic information. It can be understood that as long as the real-time broadcast content does not involve illegal characteristic information, it can be directly determined to be normal broadcast content. Broadcast content with illegal characteristic information needs to be combined with supplementary normal review content for special situation analysis and judgment.

[0052] If there is a combination of clustering basic information and remaining necessary information under any clustering violation characteristic information in the real-time broadcast content, a review screening analysis is performed in combination with the corresponding type-supplemented normal review content to form a supplementary review screening result, including: determining the combination of clustering basic information and remaining necessary information in the real-time broadcast content, and traversing the type-supplemented normal review content: if there is normal information in the real-time broadcast content that is the same as the type-supplemented normal review content, and contains a combination of clustering basic information and remaining necessary information, the corresponding real-time broadcast content is retained and marked as review-screened real-time broadcast content; if there is no normal information in the real-time broadcast content that is the same as the type-supplemented normal review content, the corresponding real-time broadcast content is deleted.

[0053] For real-time broadcasting content with information that contains illegal characteristics, when using the supplementary normal review content to make judgments, the main thing is to determine whether the expanded amount of illegal information in the real-time broadcasting content exists in the supplementary normal review content. If it does, it can be determined that although the real-time broadcasting content contains illegal characteristics, it does not meet the requirements of violations overall.

[0054] S5: Collect, review and screen the comment information of the real-time broadcast content, and perform additional screening based on the feature data of the missed comments to form a complete screening of the real-time broadcast content.

[0055] Collect, review and filter comment information on real-time broadcast content, and perform additional screening based on feature data of missed comments to form complete screened real-time broadcast content, including: collect, review and filter comment information on real-time broadcast content, and perform vocabulary decomposition on the comment information to form a decomposition vocabulary set; perform additional screening and analysis based on the decomposition vocabulary set and in combination with feature data of missed comments of corresponding content types to determine complete screened real-time broadcast content.

[0056] Real-time broadcast content that has been screened for illegal feature information needs to be properly monitored to avoid missing illegal content due to algorithm defects. Screening mainly involves continuously obtaining comment information, then breaking down the vocabulary, and comparing and analyzing it with the feature data of missed comments.

[0057] Based on the disassembled vocabulary set and combined with the characteristic data of missed comments of corresponding content types, supplementary screening and analysis are carried out to determine the complete screening of real-time broadcast content, including: if the review information of the screened real-time broadcast content contains all the repetitive words of any missed comment repetitive vocabulary group, and does not meet the requirement that the repetition amount of all repetitive words reaches the corresponding repetitive vocabulary amount, it is marked as the primary missed broadcast content; if the review information of the screened real-time broadcast content does not contain all the repetitive words of any missed comment repetitive vocabulary group, but some of the repetitive words it has meet the requirement that the repetition amount of all repetitive words reaches the corresponding repetitive vocabulary amount, it is marked as important missed broadcast content; if the review information of the screened real-time broadcast content does not contain all the repetitive words of any missed comment repetitive vocabulary group, and some of the repetitive words it has do not meet the requirement that the repetition amount of all repetitive words reaches the corresponding repetitive vocabulary amount, it is marked as secondary missed broadcast content; if the review information of the screened real-time broadcast content does not contain any repetitive words of any missed comment repetitive vocabulary group, it is marked as normal real-time broadcast content.

[0058] Since the feature analysis of missed comments is a continuous process, the analysis is to calibrate and continuously observe different comparison situations. Of course, for the comment information that has never had any missed comment feature information, it can be completely determined that the corresponding real-time broadcast content is normal. For the comment data with missed comment feature information, different feature occurrences are considered for classification and calibration, so as to provide a distinction between key and secondary tasks for subsequent continuous observation, so as to achieve rational use of resources.

[0059] In summary, the broadcast content screening method for content review provided by the embodiment of the present invention has the following beneficial effects:

[0060] This method collects historical review data of broadcast content, establishes a database for feature extraction, and extracts feature information based on different types of broadcast content. Compared with direct feature information extraction, this method can greatly reduce the scope of feature information collection. On the one hand, it can improve the accuracy of feature information extraction, which is conducive to improving the accuracy of subsequent content screening. On the other hand, because it is not based on type, it also greatly reduces the efficiency of subsequent screening and analysis. At the same time, in view of the missed reviews caused by the imperfection of intelligent algorithms, the missed review content screening feature information based on comment information is established, which is conducive to the subsequent supplement of intelligent screening, greatly improving the efficiency and accuracy of broadcast content screening, and fully ensuring the rationality of content screening.

[0061] In the embodiments of the present application, "indication" may include direct indication and indirect indication, and may also include explicit indication and implicit indication. The information indicated by a certain information is called information to be indicated. In the specific implementation process, there are many ways to indicate the information to be indicated, such as but not limited to, the information to be indicated can be directly indicated, such as the information to be indicated itself or the index of the information to be indicated. The information to be indicated can also be indirectly indicated by indicating other information, wherein there is an association between the other information and the information to be indicated. It is also possible to indicate only a part of the information to be indicated, while the other parts of the information to be indicated are known or agreed in advance. For example, the indication of specific information can also be achieved by means of the arrangement order of each piece of information agreed in advance (for example, specified by the protocol), thereby reducing the indication overhead to a certain extent. At the same time, the common parts of each piece of information can also be identified and indicated uniformly to reduce the indication overhead caused by indicating the same information separately.

[0062] In addition, the specific indication method may also be various existing indication methods, such as but not limited to the above-mentioned indication methods and various combinations thereof. The specific details of the various indication methods can refer to the prior art and will not be repeated herein. As can be seen from the above, for example, when it is necessary to indicate multiple information of the same type, different indication methods may be used for different information. In the specific implementation process, the desired indication method can be selected according to specific needs. The embodiment of the present application does not limit the selected indication method. In this way, the indication method involved in the embodiment of the present application should be understood to cover various methods that can enable the party to be indicated to obtain the information to be indicated.

[0063] It should be understood that the information to be indicated can be sent as a whole, or divided into multiple sub-information and sent separately, and the sending period and / or sending time of these sub-information can be the same or different. The specific sending method is not limited in the embodiment of the present application. Among them, the sending period and / or sending time of these sub-information can be pre-defined, for example, pre-defined according to a protocol, or can be configured by the sending end device by sending configuration information to the receiving end device.

[0064] "Pre-definition" or "pre-configuration" can be implemented by pre-saving corresponding codes, tables or other methods that can be used to indicate relevant information in the device, and the embodiments of the present application do not limit the specific implementation method. Among them, "saving" can mean saving in one or more memories. The one or more memories can be set separately or integrated in an encoder or decoder, a processor, or a communication device. The one or more memories can also be partially set separately and partially integrated in a decoder, a processor, or a communication device. The type of memory can be any form of storage medium, which is not limited by the embodiments of the present application.

[0065] The "protocol" involved in the embodiments of the present application may refer to a protocol family in the communication field, a standard protocol with a similar protocol family frame structure, or a related protocol used in future communication systems, and the embodiments of the present application do not make specific limitations on this.

[0066] In the embodiments of the present application, descriptions such as "when...", "in the case of...", "if" and "if" all mean that the device will make corresponding processing under certain objective circumstances. It does not limit the time, nor does it require the device to have a judgment action when implementing it, nor does it mean that there are other limitations.

[0067] In the description of the embodiments of the present application, unless otherwise specified, " / " indicates that the objects associated before and after are in an "or" relationship. For example, A / B can represent A or B; "and / or" in the embodiments of the present application is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. These three situations, where A and B can be singular or plural. In addition, in the description of the embodiments of the present application, unless otherwise specified, "multiple" refers to two or more than two. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple. In addition, in order to facilitate the clear description of the technical solution of the embodiments of the present application, in the embodiments of the present application, the words "first" and "second" are used to distinguish the same or similar items with basically the same functions and effects. Those skilled in the art will appreciate that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit the difference. At the same time, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way for easy understanding.

[0068] It should be understood that the processor in the embodiments of the present application may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0069] It should also be understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0070] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0071] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0072] In this application, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0073] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0074] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0075] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0076] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0077] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0078] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0079] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0080] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for screening broadcast content for content review, characterized in that: include: Collect historical broadcast content review data and establish a broadcast content review database; Extracting historical illegal broadcast contents and historical normal broadcast contents of different content types from the broadcast content history review database, performing feature extraction based on content types, and forming content type broadcast review feature information; Extracting historical missed-review broadcast contents of different content types from the broadcast content historical review database, extracting feature information based on comments, and forming missed-review comment feature data; Acquire real-time broadcast content, and perform review and screening according to the broadcast review characteristic information of the content type to form review and screen real-time broadcast content; Collecting the review information of the screened real-time broadcast content, and performing supplementary screening based on the missed review feature data to form a complete screened real-time broadcast content; The extracting of historical illegal broadcast contents and historical normal broadcast contents of different content types in the broadcast content history review database, performing feature extraction based on content types, and forming content type broadcast review feature information includes: Extract the historical illegal broadcast content and the historical normal broadcast content of the same content type from the broadcast content history review database to form type characteristic basic content data corresponding to the content type; Extracting illegal feature information from all the historical illegal broadcast contents in the type feature basic content data to form type illegal content review feature information; According to the type illegal content review feature information, extracting supplementary normal feature information of all the historical normal broadcast contents in the type feature basic content data to form type supplementary normal content review feature information; The review characteristic information of the illegal content of the type and the review characteristic information of the supplementary normal content of the type are collected to form the broadcast review characteristic information of the content type.

2. The method for screening broadcast content for content review according to claim 1, characterized in that: The step of extracting illegal feature information from all the historical illegal broadcast contents in the type feature basic content data to form type illegal content review feature information includes: Extracting violation information of different historical illegal broadcast contents from the type feature basic content data to form type initial violation information; The initial violation information of different types is compared and clustered in the following way: If there is the same information between the initial violation information of different types, extract the same part of the information to determine it as clustering basic information, obtain the remaining necessary information in the initial violation information of different types except the clustering basic information to form a remaining necessary information set, and combine the clustering basic information and the corresponding remaining necessary information set to form clustering violation feature information; The different clustered violation feature information are gathered to form the type of violation content review feature information corresponding to the content type.

3. The method for screening broadcast content for content review according to claim 2, characterized in that: The method of extracting supplementary normal feature information from all the historical normal broadcast contents in the type feature basic content data according to the type illegal content review feature information to form type supplementary normal content review feature information includes: The following additional normal feature information is extracted for the different clustered illegal feature information of the type illegal content review feature information: If normal information including the type of initial violation information exists in the historical normal broadcast content, the normal information with information amount exceeding that corresponding to the type of initial violation information is extracted to form the type of supplementary normal review content under the type of violation content review characteristic information corresponding to the type of initial violation information; The different types of supplementary normal review contents under the type of illegal content review characteristic information are gathered to form corresponding type of supplementary normal content review characteristic information.

4. The method for screening broadcast content for content review according to claim 3, characterized in that: The extracting of historical missed-review broadcast contents of different content types from the broadcast content historical review database, extracting feature information based on comments, and forming missed-review comment feature data includes: For different historically missed broadcast contents of the same content type, extract corresponding comment information to perform word repetition statistics, cluster the repetitive words, and form a type repetitive word data set; Determining repeated vocabulary groups of missed comments in different historical missed broadcast contents according to the type repeated vocabulary data set; Performing repetition statistics on different repetitive vocabulary groups of missed comments corresponding to the historical missed broadcast content, and determining the repetitive vocabulary amounts corresponding to different repetitive vocabulary in the repetitive vocabulary groups of missed comments; The feature information of the missed review comments is extracted according to different repetitive vocabulary groups of the missed review comments and the repetitive vocabulary amounts of different repetitive vocabulary corresponding to the repetitive vocabulary groups of the missed review comments to form the feature data of the missed review comments.

5. The method for screening broadcast content for content review according to claim 4, characterized in that: The step of extracting the feature information of missed reviews according to different repeated vocabulary groups of missed reviews and repeated vocabulary amounts of different repeated vocabulary corresponding to the repeated vocabulary groups of missed reviews to form the feature data of missed reviews includes: If different repetitive vocabulary groups of missed reviews have an inclusion relationship, the repetitive vocabulary group of missed reviews with the least number of repetitive vocabulary is determined as the missed reviews feature group, and the maximum repetitive amount corresponding to each repetitive vocabulary is determined as the missed reviews vocabulary feature amount; If the missed review review repetitive vocabulary group does not have an inclusion relationship with other missed review review repetitive vocabulary groups, the missed review review repetitive vocabulary group is determined as a missed review review feature group, and the repetition amount of each corresponding repetitive vocabulary is determined as the missed review review vocabulary feature amount; The missed review comment feature data is formed by combining different missed review comment feature groups and the missed review comment lexical feature quantities corresponding to different repetitive words in the missed review comment feature groups.

6. The method for screening broadcast content for content review according to claim 5, characterized in that: The acquiring of real-time broadcast content and screening of the content type broadcast review feature information to form the screened real-time broadcast content includes: The real-time broadcast content is obtained, and the corresponding illegal content review feature information of the type is extracted, and the review and screening is performed in the following manner: If the combination of the cluster basic information and the remaining necessary information under any cluster violation characteristic information does not exist in the real-time broadcast content, the corresponding real-time broadcast content is retained and marked as the real-time broadcast content for review and screening; If any combination of the clustering basic information and the remaining necessary information under the clustering violation characteristic information exists in the real-time broadcast content, the corresponding type of supplementary normal review content is combined for review, screening and analysis to form a supplementary review and screening result.

7. The method for screening broadcast content for content review according to claim 6, characterized in that: If the real-time broadcast content contains any combination of the cluster basic information and the remaining necessary information under the cluster violation characteristic information, the corresponding type of supplementary normal review content is combined for review screening analysis to form a supplementary review screening result, including: Determine the combination of the clustering basic information and the remaining necessary information present in the real-time broadcast content, and traverse in the type-supplemented normal review content: If the real-time broadcast content contains the same normal information as that in the type of supplementary normal review content, and contains a combination of the clustering basic information and the remaining necessary information, the corresponding real-time broadcast content is retained and marked as the review and screening real-time broadcast content; If the real-time broadcast content does not contain the same normal information as the type of supplementary normal review content, the corresponding real-time broadcast content is deleted.

8. The method for screening broadcast content for content review according to claim 7, characterized in that: The collecting of the review information of the real-time broadcast content and the supplementary screening according to the feature data of the missed reviews to form a complete screening of the real-time broadcast content include: Collecting the review information of the real-time broadcast content, and performing vocabulary decomposition on the review information to form a decomposition vocabulary set; A supplementary screening analysis is performed based on the disassembled vocabulary set and in combination with the missed review feature data of the corresponding content type to determine the complete screening of the real-time broadcast content.

9. The method for screening broadcast content for content review according to claim 8, characterized in that: The method of performing supplementary screening and analysis based on the disassembled vocabulary set and combining the missed review feature data of the corresponding content type to determine the complete screening of real-time broadcast content includes: If the review information of the real-time broadcast content being screened contains all the repetitive words in any of the repetitive word groups of missed reviews, and the repetitive amount of all the repetitive words does not meet the requirement that the repetitive amount of all the repetitive words reaches the corresponding repetitive word amount, then it is marked as the primary missed review broadcast content; If the review information of the real-time broadcast content does not contain all the repetitive words in any of the repetitive word groups of the missed review comments, but has some repetitive words that satisfy the repetition amount of all the repetitive words reaching the corresponding repetitive word amount, it is marked as important missed review broadcast content; If the review information of the real-time broadcast content does not contain all the repetitive words in any of the repetitive word groups of the missed review comments, and the repetitive words of some of the existing repetitive words do not meet the requirement that the repetitive amount of all the repetitive words reaches the corresponding repetitive word amount, it is marked as secondary missed review broadcast content; If the review information of the real-time broadcast content screened does not contain any repetitive words in the repetitive word group of missed comments, it is marked as normal real-time broadcast content.

Images