Multi-channel vocabulary filtering method and device

Abstract

The invention relates to the technical field of keyword filtering, and discloses a multi-channel vocabulary filtering method. Obtaining filtering system target information or filtering system requirement information, and generating a filtering word library according to the filtering system target information or the filtering system requirement information; acquiring sample data, and preprocessing the sample data; selecting a natural language processing model and an image / video recognition model according to the filtering system target information or the filtering system requirement information, and training the natural language processing model and the image / video recognition model by using the sample data; obtaining to-be-detected text information, and performing context and semantic analysis on the to-be-detected text information through the natural language processing model to generate analysis result information; sending the analysis result information to a manual auditing terminal, and obtaining auditing result information of the manual auditing terminal; obtaining filtering vocabulary updating information, and updating the filtering word library according to the filtering vocabulary updating information; the method has the advantages of low false alarm rate and high accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of keyword filtering, and in particular to a multi-channel vocabulary filtering method and device. Background Art

[0002] Keyword filtering, as a means of network content management, does have some defects, deficiencies or problems in actual application, mainly including the following aspects: 1. High false positive and false negative rates: Keyword filtering mainly relies on matching and identifying specific keywords, but this method is often too simple and mechanical, which can easily lead to false positives and false negatives. For example, some normal, non-violation texts may be misjudged as violation content because they contain words similar to violation keywords, while some carefully constructed violation content may bypass keyword filtering through homophones, special symbols, etc.

[0003] 2. Unable to understand context: Keyword filtering can usually only match and identify single keywords, but cannot understand the overall content and context of the text. Therefore, it is difficult to accurately determine whether a text is truly illegal. For example, a text containing the keyword "death" may be discussing a tragic event or describing a fictional story, but keyword filtering cannot distinguish between the two situations.

[0004] 3. Unable to cope with language changes: Internet language changes rapidly, with new words and expressions emerging in an endless stream. The keyword library for keyword filtering often needs to be updated and maintained regularly, otherwise it may not be able to identify new illegal content. However, due to the complexity and diversity of Internet language, the update of the keyword library often cannot keep up with the speed of language change.

[0005] 4. Easy to be bypassed: Since the mechanism and principle of keyword filtering are relatively simple, some users may take some technical means to bypass keyword filtering. For example, they may use homophones, special symbols, pictures, etc. to transmit illegal information, thereby bypassing the detection of keyword filtering.

[0006] 5. Difficulty in balancing censorship and free expression: Keyword filtering can limit the spread of illegal content to a certain extent, but it may also impose certain restrictions on users' free expression. For example, some normal, non-illegal speech may be misjudged as illegal content because it contains words similar to illegal keywords, thus limiting users' free expression.

[0007] Therefore, a multi-channel vocabulary filtering method and device are provided to solve the above problems. Summary of the invention

[0008] The main purpose of the present invention is to solve the problems of existing keyword filtering, such as high false positive and false negative rates, inability to understand context, inability to cope with language changes, easy bypassing, and difficulty in balancing censorship and free expression.

[0009] A first aspect of the present invention provides a multi-channel vocabulary filtering method, the multi-channel vocabulary filtering method comprising: Acquire filtering system target information or filtering system requirement information, and generate a filtering vocabulary according to the filtering system target information or filtering system requirement information; Obtain sample data and preprocess the sample data; Select a natural language processing model and an image / video recognition model according to the target information of the filtering system or the requirement information of the filtering system, and train the natural language processing model and the image / video recognition model using the sample data; Obtain the text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; Send the analysis result information to the manual review end to obtain the review result information of the manual review end; Obtain filter vocabulary update information, and update the filter vocabulary library according to the filter vocabulary update information.

[0010] Optionally, the obtaining sample data and preprocessing the sample data includes: Acquire sample data, wherein the sample data includes samples containing filtered vocabulary and samples not containing filtered vocabulary; The sample data is cleaned, labeled and formatted in turn.

[0011] Optionally, the obtaining of the text information to be detected, performing context and semantic analysis on the text information to be detected by a natural language processing model, and generating analysis result information includes: The text information to be detected is obtained, and a natural language processing model is used to perform context analysis on the text information to be detected to generate analysis results, wherein the analysis results include the overall intention information and deep meaning information of the text.

[0012] Optionally, the acquiring filtering vocabulary update information and updating the filtering vocabulary library according to the filtering vocabulary update information includes: Monitor network content and collect new filtering words and expressions; Update the filter word library according to new filter words and filter expressions; Obtain user feedback information and update the filter vocabulary based on the user feedback information.

[0013] Optionally, also include: Obtaining user behavior data information, wherein the user behavior data information includes posting frequency information and reply behavior information; Determine whether the user is a potential filtering user based on the user behavior data information; if so, conduct a strict filtering review on the user; if not, conduct a normal filtering review on the user.

[0014] Optionally, judging whether the user is a potential filtering user according to the user behavior data information includes: Extract key word information from user behavior data, use TF-IDF to evaluate the importance of key word information, and sort the key word information from high to low according to the importance rating; Extract sentence information containing key word information in descending order according to the importance, and generate sentence meaning information from the sentence information using a neural network model; Determine whether the user is a potential filtering user based on the sentence meaning information.

[0015] Optionally, the sending of the analysis result information to the manual review end and obtaining the review result information of the manual review end includes: The manual review end receives the analysis result information and obtains the appeal information at the same time; The manual review end generates secondary review result information based on the analysis result information and appeal information.

[0016] A second aspect of the present invention provides a multi-channel vocabulary filtering device, comprising: A filtering word library generation module is used to obtain filtering system target information or filtering system requirement information, and generate a filtering word library according to the filtering system target information or filtering system requirement information; A sample data collection module is used to obtain sample data and pre-process the sample data; A model selection training module is used to select a natural language processing model and an image / video recognition model according to the target information of the filtering system or the requirement information of the filtering system, and train the natural language processing model and the image / video recognition model using sample data; A model analysis module is used to obtain text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; The manual review module is used to send the analysis result information to the manual review end and obtain the review result information of the manual review end; The filtering vocabulary updating module is used to obtain filtering vocabulary updating information and update the filtering vocabulary according to the filtering vocabulary updating information.

[0017] A third aspect of the present invention provides an electronic device, the electronic device comprising a memory and at least one processor, the memory storing instructions; The at least one processor calls the instructions in the memory to enable the electronic device to execute the various steps of the multi-channel vocabulary filtering method as described above.

[0018] A fourth aspect of the present invention provides a computer-readable storage medium having instructions stored thereon, and when the instructions are executed by a processor, the various steps of the multi-channel vocabulary filtering method described above are implemented.

[0019] The technical solution of the present invention provides a multi-channel vocabulary filtering method, which introduces a natural language processing model to perform context analysis, focusing not only on the keywords themselves but also on the contextual environment in which the keywords appear, greatly reducing the false alarm rate and improving the accuracy of filtering. It can also automatically update and adjust the keyword thesaurus, reduce the false alarm rate, increase manual review, and ensure the accuracy and fairness of the filtering results. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flowchart of a multi-channel vocabulary filtering method provided by a first embodiment of the present invention; Figure 2 A flowchart of a multi-channel vocabulary filtering method provided by a second embodiment of the present invention; Figure 3 A flowchart of a multi-channel vocabulary filtering method provided by a third embodiment of the present invention; Figure 4 A schematic diagram of the structure of a multi-channel vocabulary filtering device provided by an embodiment of the present invention; Figure 5 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The embodiment of the present invention provides a multi-channel vocabulary filtering method, including obtaining filtering system target information or filtering system requirement information, generating a filtering vocabulary library according to the filtering system target information or filtering system requirement information; obtaining sample data, and preprocessing the sample data; selecting a natural language processing model and an image / video recognition model according to the filtering system target information or filtering system requirement information, and using the sample data to train the natural language processing model and the image / video recognition model; obtaining text information to be detected, performing context and semantic analysis on the text information to be detected through a natural language processing model, and generating analysis result information; sending the analysis result information to a manual review end, and obtaining review result information of the manual review end; obtaining filtering vocabulary update information, and updating the filtering vocabulary library according to the filtering vocabulary update information. The present invention solves the problems of existing keyword filtering, such as high false positive and false negative rates, inability to understand context, inability to cope with language changes, easy bypass, and difficulty in balancing review and free expression.

[0022] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 The first embodiment of the multi-channel vocabulary filtering method in the embodiment of the present invention includes: Acquire filtering system target information or filtering system requirement information, and generate a filtering vocabulary according to the filtering system target information or filtering system requirement information; Obtain sample data and preprocess the sample data; Specifically, they include: Acquire sample data, the sample data including samples containing filtered vocabulary and samples not containing filtered vocabulary; perform data cleaning, data labeling and data formatting on the sample data in sequence; preprocess the sample data to facilitate subsequent model training and evaluation; Select a natural language processing model and an image / video recognition model according to the target information or the requirement information of the filtering system, and use sample data to train the natural language processing model and the image / video recognition model; by using the natural language processing model, semantics and context can be processed to reduce false positives and false negatives; Obtain the text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; Specifically, it includes: obtaining text information to be detected, performing context analysis on the text information to be detected using a natural language processing model, and generating analysis results, wherein the analysis results include information on the overall intention of the text and information on its deep meaning; Send the analysis result information to the manual review end to obtain the review result information of the manual review end; Obtaining filtering vocabulary update information, and updating the filtering vocabulary library according to the filtering vocabulary update information; Specifically, they include: Monitor network content and collect new filtering words and expressions; Update the filter word library according to new filter words and filter expressions; Obtain user feedback information and update the filter word library based on the user feedback information. The filter word library can be updated in real time to improve the accuracy of filtering.

[0024] See also Figure 2 The second embodiment of the multi-channel vocabulary filtering method in the embodiment of the present invention includes: Acquire filtering system target information or filtering system requirement information, and generate a filtering vocabulary according to the filtering system target information or filtering system requirement information; Obtain sample data and preprocess the sample data; Specifically, they include: Acquire sample data, the sample data including samples containing filtered vocabulary and samples not containing filtered vocabulary; perform data cleaning, data labeling and data formatting on the sample data in sequence; preprocess the sample data to facilitate subsequent model training and evaluation; Select a natural language processing model and an image / video recognition model according to the target information or the requirement information of the filtering system, and use sample data to train the natural language processing model and the image / video recognition model; by using the natural language processing model, semantics and context can be processed to reduce false positives and false negatives; Obtain the text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; Specifically, it includes: obtaining text information to be detected, performing context analysis on the text information to be detected using a natural language processing model, and generating analysis results, wherein the analysis results include information on the overall intention of the text and information on its deep meaning; Send the analysis result information to the manual review end to obtain the review result information of the manual review end; Obtaining filtering vocabulary update information, and updating the filtering vocabulary library according to the filtering vocabulary update information; Specifically, they include: Monitor network content and collect new filtering words and expressions; Update the filter word library according to new filter words and filter expressions; Obtain user feedback information and update the filter word library based on the user feedback information. The filter word library can be updated in real time to improve the accuracy of filtering.

[0025] It also includes: obtaining user behavior data information, the user behavior data information including posting frequency information and reply behavior information; judging whether the user is a potential filtering user according to the user behavior data information; if so, performing a strict filtering review on the user; if not, performing a normal filtering review on the user.

[0026] See also Figure 3 The third embodiment of the multi-channel vocabulary filtering method in the embodiment of the present invention includes: Acquire filtering system target information or filtering system requirement information, and generate a filtering vocabulary according to the filtering system target information or filtering system requirement information; Obtain sample data and preprocess the sample data; Specifically, they include: Acquire sample data, the sample data including samples containing filtered vocabulary and samples not containing filtered vocabulary; perform data cleaning, data labeling and data formatting on the sample data in sequence; preprocess the sample data to facilitate subsequent model training and evaluation; Select a natural language processing model and an image / video recognition model according to the target information or the requirement information of the filtering system, and use sample data to train the natural language processing model and the image / video recognition model; by using the natural language processing model, semantics and context can be processed to reduce false positives and false negatives; Obtain the text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; Specifically, it includes: obtaining text information to be detected, performing context analysis on the text information to be detected using a natural language processing model, and generating analysis results, wherein the analysis results include information on the overall intention of the text and information on its deep meaning; Send the analysis result information to the manual review end to obtain the review result information of the manual review end; Specifically include: The manual review end receives the analysis result information and obtains the appeal information at the same time; The manual review end generates secondary review result information based on the analysis result information and appeal information.

[0027] Obtaining filtering vocabulary update information, and updating the filtering vocabulary library according to the filtering vocabulary update information; Specifically, they include: Monitor network content and collect new filtering words and expressions; Update the filter word library according to new filter words and filter expressions; Obtain user feedback information and update the filter word library based on the user feedback information. The filter word library can be updated in real time to improve the accuracy of filtering.

[0028] It also includes: obtaining user behavior data information, the user behavior data information including posting frequency information and reply behavior information; judging whether the user is a potential filtering user according to the user behavior data information; if so, performing a strict filtering review on the user; if not, performing a normal filtering review on the user.

[0029] See also Figure 3 The fourth embodiment of the multi-channel vocabulary filtering method in the embodiment of the present invention includes: Acquire filtering system target information or filtering system requirement information, and generate a filtering vocabulary according to the filtering system target information or filtering system requirement information; Obtain sample data and preprocess the sample data; Specifically, they include: Acquire sample data, the sample data including samples containing filtered vocabulary and samples not containing filtered vocabulary; perform data cleaning, data labeling and data formatting on the sample data in sequence; preprocess the sample data to facilitate subsequent model training and evaluation; Select a natural language processing model and an image / video recognition model according to the target information or the requirement information of the filtering system, and use sample data to train the natural language processing model and the image / video recognition model; by using the natural language processing model, semantics and context can be processed to reduce false positives and false negatives; Obtain the text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; Specifically, it includes: obtaining text information to be detected, performing context analysis on the text information to be detected using a natural language processing model, and generating analysis results, wherein the analysis results include information on the overall intention of the text and information on its deep meaning; Send the analysis result information to the manual review end to obtain the review result information of the manual review end; Specifically include: The manual review end receives the analysis result information and obtains the appeal information at the same time; The manual review end generates secondary review result information based on the analysis result information and appeal information.

[0030] Obtaining filtering vocabulary update information, and updating the filtering vocabulary library according to the filtering vocabulary update information; Specifically, they include: Monitor network content and collect new filtering words and expressions; Update the filter word library according to new filter words and filter expressions; Obtain user feedback information and update the filter word library based on the user feedback information. The filter word library can be updated in real time to improve the accuracy of filtering.

[0031] The method further includes: obtaining user behavior data information, wherein the user behavior data information includes posting frequency information and reply behavior information; judging whether the user is a potential filtering user according to the user behavior data information; if so, performing a strict filtering review on the user; if not, performing a normal filtering review on the user; Judging whether a user is a potential filtering user based on user behavior data information includes: Extract key word information from the reply behavior information in the user behavior data information, use TF-IDF (TermFrequency-Inverse Document Frequency) to evaluate the importance of key word information, and arrange the key word information from high to low according to the importance rating; Extract sentence information containing key word information in descending order according to the importance, and generate sentence meaning information from the sentence information using a neural network model; Determine whether the user is a potential filtering user based on the sentence meaning information.

[0032] See also Figure 3 The fifth embodiment of the multi-channel vocabulary filtering method in the embodiment of the present invention includes: Acquire filtering system target information or filtering system requirement information, and generate a filtering vocabulary according to the filtering system target information or filtering system requirement information; Obtain sample data and preprocess the sample data; Specifically, they include: Acquire sample data, the sample data including samples containing filtered vocabulary and samples not containing filtered vocabulary; perform data cleaning, data labeling and data formatting on the sample data in sequence; preprocess the sample data to facilitate subsequent model training and evaluation; Select a natural language processing model and an image / video recognition model according to the target information or the requirement information of the filtering system, and use sample data to train the natural language processing model and the image / video recognition model; by using the natural language processing model, semantics and context can be processed to reduce false positives and false negatives; Obtain the text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; Specifically, it includes: obtaining text information to be detected, performing context analysis on the text information to be detected using a natural language processing model, and generating analysis results, wherein the analysis results include information on the overall intention of the text and information on its deep meaning; Send the analysis result information to the manual review end to obtain the review result information of the manual review end; Specifically include: The manual review end receives the analysis result information and obtains the appeal information at the same time; The manual review end generates secondary review result information based on the analysis result information and appeal information.

[0033] Obtaining filtering vocabulary update information, and updating the filtering vocabulary library according to the filtering vocabulary update information; Specifically, they include: Monitor network content and collect new filtering words and expressions; Update the filter word library according to new filter words and filter expressions; Obtain user feedback information and update the filter word library based on the user feedback information. The filter word library can be updated in real time to improve the accuracy of filtering.

[0034] The method further includes: obtaining user behavior data information, wherein the user behavior data information includes posting frequency information and reply behavior information; judging whether the user is a potential filtering user according to the user behavior data information; if so, performing a strict filtering review on the user; if not, performing a normal filtering review on the user; Judging whether a user is a potential filtering user based on user behavior data information includes: Extract key word information from the reply behavior information in the user behavior data information, use TF-IDF (TermFrequency-Inverse Document Frequency) to evaluate the importance of key word information, and arrange the key word information from high to low according to the importance rating; Extract sentence information containing key word information in descending order according to the importance, and generate sentence meaning information from the sentence information using a neural network model; Judging whether the user is a potential filtering user according to the sentence meaning information, the judging standard may be that if the sentence meaning information ranked before the preset ranking (e.g., the top three, top four, or top ten) in terms of importance all contain filtering words or the meaning of filtering words, then the user is judged to be a potential filtering user; otherwise, the user is judged to be a normal user; The judgment criteria can also be that if more than a preset percentage of the sentence meaning information in the importance ranking before the preset ranking contains filtered words or the meaning of filtered words, then the user is judged to be a potential filtered user; if not, then the user is judged to be a normal user.

[0035] The above describes the multi-channel vocabulary filtering method in the embodiment of the present invention. The following describes the multi-channel vocabulary filtering device in the embodiment of the present invention. Figure 4 In the embodiment of the present invention, the multi-channel vocabulary filtering device for the above embodiment includes: A filtering word library generation module 401 is used to obtain filtering system target information or filtering system requirement information, and generate a filtering word library according to the filtering system target information or filtering system requirement information; The sample data collection module 402 is used to obtain sample data and pre-process the sample data; A model selection and training module 403 is used to select a natural language processing model and an image / video recognition model according to the filtering system target information or the filtering system requirement information, and train the natural language processing model and the image / video recognition model using sample data; Model analysis module 404, used to obtain text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; The manual review module 405 is used to send the analysis result information to the manual review end and obtain the review result information of the manual review end; The filtering vocabulary updating module 406 is used to obtain filtering vocabulary update information and update the filtering vocabulary according to the filtering vocabulary update information.

[0036] above Figure 4The multi-channel vocabulary filtering device in the embodiment of the present invention is described in detail from the perspective of modular functional entities, and the electronic device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0037] Figure 5 7 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device 700 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 710 (for example, one or more processors) and a memory 720, and one or more storage media 730 (for example, one or more mass storage devices) storing application programs 733 or data 732. Among them, the memory 720 and the storage medium 730 can be temporary storage or permanent storage. The program stored in the storage medium 730 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the electronic device 700. Furthermore, the processor 710 may be configured to communicate with the storage medium 730 to execute a series of instruction operations in the storage medium 730 on the electronic device 700.

[0038] The electronic device 700 may also include one or more power supplies 740, one or more wired or wireless network interfaces 750, one or more input and output interfaces 750, and / or one or more operating systems 731, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 5 The structure of the electronic device shown does not constitute a limitation on the electronic device, and may include more or less components than shown in the figure, or combine some components, or arrange the components differently.

[0039] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of a multi-channel vocabulary filtering method.

[0040] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0041] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. 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 code.

[0042] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-channel vocabulary filtering method, characterized in that: The multi-channel vocabulary filtering method includes: Acquire filtering system target information or filtering system requirement information, and generate a filtering vocabulary according to the filtering system target information or filtering system requirement information; Obtain sample data and preprocess the sample data; Select a natural language processing model and an image / video recognition model according to the target information of the filtering system or the requirement information of the filtering system, and train the natural language processing model and the image / video recognition model using the sample data; Obtain the text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; Send the analysis result information to the manual review end to obtain the review result information of the manual review end; Obtain filter vocabulary update information, and update the filter vocabulary library according to the filter vocabulary update information.

2. The multi-channel vocabulary filtering method according to claim 1, characterized in that: The obtaining of sample data and preprocessing of the sample data comprises: Acquire sample data, wherein the sample data includes samples containing filtered vocabulary and samples not containing filtered vocabulary; The sample data is cleaned, labeled and formatted in turn.

3. The multi-channel vocabulary filtering method according to claim 1, characterized in that: The obtaining of the text information to be detected, performing context and semantic analysis on the text information to be detected by using a natural language processing model, and generating analysis result information includes: The text information to be detected is obtained, and a natural language processing model is used to perform context analysis on the text information to be detected to generate analysis results, wherein the analysis results include the overall intention information and deep meaning information of the text.

4. The multi-channel vocabulary filtering method according to claim 1, characterized in that: The obtaining of filtering vocabulary update information and updating the filtering vocabulary base according to the filtering vocabulary update information comprises: Monitor network content and collect new filtering words and expressions; Update the filter word library according to new filter words and filter expressions; Obtain user feedback information and update the filter vocabulary based on the user feedback information.

5. The multi-channel vocabulary filtering method according to claim 1, characterized in that: Also includes: Obtaining user behavior data information, wherein the user behavior data information includes posting frequency information and reply behavior information; Determine whether the user is a potential filtering user based on the user behavior data information; if so, conduct a strict filtering review on the user; if not, conduct a normal filtering review on the user.

6. The multi-channel vocabulary filtering method according to claim 5, characterized in that: Determining whether a user is a potential filtering user according to the user behavior data information includes: Extract key word information from user behavior data, use TF-IDF to evaluate the importance of key word information, and sort the key word information from high to low according to the importance rating; Extract sentence information containing key word information in descending order according to the importance, and generate sentence meaning information from the sentence information using a neural network model; Determine whether the user is a potential filtering user based on the sentence meaning information.

7. The multi-channel vocabulary filtering method according to claim 1, characterized in that: The step of sending the analysis result information to the manual review end and obtaining the review result information from the manual review end includes: The manual review end receives the analysis result information and obtains the appeal information at the same time; The manual review end generates secondary review result information based on the analysis result information and appeal information.

8. A multi-channel vocabulary filtering device, characterized in that: include: A filtering word library generation module is used to obtain filtering system target information or filtering system requirement information, and generate a filtering word library according to the filtering system target information or filtering system requirement information; A sample data collection module is used to obtain sample data and pre-process the sample data; A model selection training module is used to select a natural language processing model and an image / video recognition model according to the target information of the filtering system or the requirement information of the filtering system, and train the natural language processing model and the image / video recognition model using sample data; A model analysis module is used to obtain text information to be detected, perform context and semantic analysis on the text information to be detected through a natural language processing model, and generate analysis result information; The manual review module is used to send the analysis result information to the manual review end and obtain the review result information of the manual review end; The filtering vocabulary updating module is used to obtain filtering vocabulary updating information and update the filtering vocabulary according to the filtering vocabulary updating information.

9. An electronic device, comprising a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the electronic device to execute each step of the multi-channel vocabulary filtering method according to any one of claims 1 to 7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by a processor, the steps of the multi-channel vocabulary filtering method according to any one of claims 1 to 7 are implemented.

Images