Work content auditing system, method and equipment and storage medium
By using AI agents to generate virtual detectors with multiple detection dimensions and feature extraction from the work description extraction module, the reusability and accuracy issues of existing systems when review rules change are resolved, thus achieving efficient work content review.
Patent Information
- Application Number
- CN202511067097.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-31
- Publication Date
- 2025-11-07
AI Technical Summary
The existing work review system cannot perform reviews properly when the event theme and review rules change, resulting in reduced reusability and review accuracy.
An AI agent is used as the virtual detector generation module. Based on the received review rule text, a virtual detector with multiple detection dimensions is generated. The work description extraction module performs feature extraction and interference data removal. Combined with the semantic search module, the information retrieval efficiency and review accuracy are improved.
This system improves the reusability and accuracy of the content review system without requiring human intervention, reduces the difficulty of reviewing with virtual detectors, and enhances the accuracy of content review.
Smart Images
Figure CN120911433A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of content review, and in particular to a work content review system, method, device and storage medium. BACKGROUND
[0002] With the development and application of digital technology, it is a common practice to build a work review system in enterprise internal activities, such as allowing participants to upload works in the form of images, audio, text, etc. to the work review system, so as to facilitate subsequent content review and scoring.
[0003] However, the activity theme and review rules of different internal activities may differ greatly, and the review model used by the existing work review system is mostly built based on fixed themes and review rules. This leads to the original work review system being unable to normally perform work review in the case of great changes in the activity theme and review rules, reducing the reusability and review accuracy of the work review system. SUMMARY
[0004] In view of the above problems, the present application provides a work content review system, method, device and storage medium to achieve the purpose of improving the reusability and review accuracy of the work review system. The specific scheme is as follows:
[0005] The first aspect of the present application provides a work content review system, comprising:
[0006] a work description extraction module and a virtual detector generation module;
[0007] The work description extraction module is configured to perform description feature extraction based on at least the received work, and send the extracted work content description features to the virtual detector generation module;
[0008] The virtual detector generation module is configured to generate virtual detectors of multiple detection dimensions based on the received review rule text of the running version, and further configured to call each virtual detector to perform rule violation review on the work content description features. The virtual detector generation module is an AI agent.
[0009] In a possible implementation, the work content review system further comprises:
[0010] a semantic search module configured to perform semantic extraction on the received activity theme text of the running version, and determine a semantic route based on the semantic extraction result.
[0011] In a possible implementation, the work content review system further comprises:
[0012] The task coordination module is used to identify the information richness of the work before the work description extraction module extracts descriptive features from the received work, and to send a content supplementation instruction to the semantic search module if the information richness is lower than a preset richness threshold.
[0013] In one possible implementation, the semantic search module is further configured to, upon receiving the content supplementation instruction, search for supplementary description information based on the semantic routing, and send the supplementary description information to the work description extraction module.
[0014] In one possible implementation, the work description extraction module performs description feature extraction based at least on the received work, including:
[0015] The work description extraction module extracts the descriptive features based on the work and the supplementary descriptive information.
[0016] In one possible implementation, the content review system further includes:
[0017] The violation content labeling module is used to receive the violation review result sent by the virtual detector, which indicates that the content contains violations, to label the work with violation content based on the violation information in the violation review result, and to output the labeling result.
[0018] In one possible implementation, the content review system further includes:
[0019] An optimization module is used to send the annotation results to the virtual detector generation module so that the virtual detector generation module can adjust the parameters of each virtual detector.
[0020] A second aspect of this application provides a method for reviewing the content of a work, applicable to the work content review system described in the first aspect of this application and any possible implementation of the first aspect, the work content review method comprising:
[0021] The text of the review rules for obtaining works and running versions;
[0022] The audit rule text is input into the virtual detector generation module, so that the virtual detector generation module generates virtual detectors with multiple detection dimensions. The virtual detector generation module is an AI agent.
[0023] The work is at least input into the work description extraction module so that the work description extraction module can extract the description features of the work and obtain the work content description features;
[0024] The work content description features are input into the virtual detector generation module, so that the virtual detector generation module invokes each virtual detector to perform violation review on the work content description features.
[0025] In a possible implementation, the work content review system further includes a semantic search module, and the work content review method further includes:
[0026] The received active topic text of the running version is input into the semantic search module, so that the semantic search module performs semantic extraction on the active topic text of the running version, and determines a semantic route based on the semantic extraction result.
[0027] In a possible implementation, the work content review system further includes a task coordination module, and the work content review method further includes:
[0028] Before the work description extraction module is used to perform description feature extraction on at least the received work, the task coordination module is used to identify information richness of the work, and in a case where the information richness is lower than a preset richness threshold, a content supplement instruction is sent to the semantic search module.
[0029] In a possible implementation, the work content review system further includes:
[0030] In a case where the content supplement instruction is received, the semantic search module is invoked to search for supplement description information based on the semantic route, and the supplement description information is sent to the work description extraction module.
[0031] In a possible implementation, the work content review system further includes:
[0032] The work description extraction module is invoked to perform the description feature extraction based on the work and the supplement description information.
[0033] In a possible implementation, the work content review system further includes a violation content labeling module, and the work content review method further includes:
[0034] The violation content labeling module is invoked to receive the violation review result sent by the virtual detector, the work is labeled with violation content based on violation information in the violation review result, and a labeling result is output.
[0035] In a possible implementation, the work content review system further includes an optimization module, and the work content review method further includes:
[0036] The optimization module is called to send the labeling result to the virtual detector generation module, so that the virtual detector generation module adjusts parameters of each virtual detector.
[0037] The third aspect of the present application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:
[0038] The memory is configured to store a computer program;
[0039] The processor is configured to execute the computer program, so that the electronic device can implement the work content auditing method of the second aspect or any implementation manner of the second aspect.
[0040] The fourth aspect of the present application provides a computer storage medium, the storage medium carries one or more computer programs, when the one or more computer programs are executed by an electronic device, the electronic device can implement the work content auditing method of the second aspect or any implementation manner of the second aspect.
[0041] Through the above technical solution, the work content auditing system, method, device and storage medium provided by the present application are provided, an AI intelligent agent is configured as a virtual detector generation module, and the virtual detector generation module generates virtual detectors of multiple detection dimensions based on the received audit rule text of the running version, so that the generated virtual detectors can be audited based on the audit rules of the running version, and manual intervention is not required in the generation process of each virtual detector, which realizes the improvement of the reusability of the work content auditing system while ensuring the auditing accuracy. Moreover, the work description extraction module is configured to perform description feature extraction based on at least the received work, and the extracted work content description features are sent to the virtual detector generation module, so that the interference data is eliminated through feature extraction, the auditing difficulty of each virtual detector is reduced, and the auditing accuracy of the work content is improved. It can be seen that the reusability and auditing accuracy of the work auditing system are improved. BRIEF DESCRIPTION OF DRAWINGS
[0042] The above and other features, advantages, and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals refer to the same or similar elements. It should be understood that the drawings are schematic, and the original and elements are not necessarily drawn according to the scale.
[0043] Figure 1 A block diagram of a work content auditing system provided by the present application is shown in the figure;
[0044] Figure 2 An architecture schematic diagram of a work description extraction module provided by the present application is shown in the figure;
[0045] Figure 3 A signaling diagram of a configuration phase of a work content review system provided by the present application is provided;
[0046] Figure 4 A signaling diagram of a running phase of a work content review system provided by the present application is provided;
[0047] Figure 5 A flowchart of a work content review method provided by the present application is provided;
[0048] Figure 6 A structural schematic diagram of an electronic device provided by the present application is provided. DETAILED DESCRIPTION
[0049] The embodiments of the present application are described below in conjunction with the accompanying drawings. The terms used in the embodiment part of the present application are only used to explain the specific embodiments of the present application, and are not intended to limit the present application.
[0050] The embodiments of the present application are described below in conjunction with the accompanying drawings. It is known to those of ordinary skill in the art that, as technology develops and new scenarios appear, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.
[0051] The terms “first”, “second”, and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged under appropriate circumstances, and this is only a distinguishing way used in the description of the embodiments of the present application to describe the objects with the same attributes. In addition, the terms “include” and “have” and any variations thereof are intended to cover non-exclusive inclusion, so that the processes, methods, systems, products or devices containing a series of units do not have to be limited to those units, but can include other units that are not clearly listed or inherent to these processes, methods, products or devices.
[0052] The first aspect of the present application provides a work content review system, as shown in Figure 1 The work content review system includes:
[0053] The work description extraction module 101 and the virtual detector generation module 102.
[0054] The work description extraction module 101 is used for feature extraction based on at least the received work, and sends the extracted work content description features to the virtual detector generation module 102.
[0055] It should be noted that in actual application scenarios, the above-mentioned work types include but are not limited to: pictures, audio, text, video, etc.
[0056] It should be noted that in actual application scenarios, the work description extraction module 101 described above can be a calling script program for calling multiple feature extraction services. It can also be a local script program after encapsulating multiple feature extraction models. As shown in Figure 2 FIG. 1 is a schematic diagram of the architecture of a work description extraction module 101 configured with three feature extraction services: picture feature extraction service, audio feature extraction service, and video feature extraction service.
[0057] It should be noted that in actual application scenarios, the work description extraction module 101 described above can be a calling script program for calling multiple feature extraction services. It can also be a local script program after encapsulating multiple feature extraction models. As shown in
[0058] Step A1, identify the suffix of the received work file. And trigger step A2.
[0059] Step A2, call the feature extraction service corresponding to the suffix identified in step A1, and use the feature extraction service to extract the description features of the work file.
[0060] In one possible implementation, the implementation of the picture feature extraction service to extract work content description features can be to extract the content of the picture using optical character recognition (OCR), and input the extracted content text into a natural language processing (NLP) algorithm to obtain text form work content description features representing the content of the picture.
[0061] In one possible implementation, the implementation of the video feature extraction service to extract work content description features can be to extract the video into multiple frames of pictures using a frame extraction algorithm, then extract the content of each frame of picture using optical character recognition (OCR), and input the extracted content text into a natural language processing (NLP) algorithm to obtain text form work content description features representing the content of the video.
[0062] In a possible implementation, the implementation manner of the audio feature extraction service extracting the work content description feature can be converting an audio file into an audio content text by using an automatic speech recognition (ASR) algorithm, and then performing semantic recognition on the audio content text by using a natural language processing (NLP) algorithm to obtain a text form of the work content description feature representing the audio content.
[0063] It should be noted that in an actual application scenario, in addition to the data representing the work content, the work also includes interference data irrelevant to the work content. Therefore, the present application extracts the description feature based on the received work at least by configuring the work description extraction module, and sends the extracted work content description feature to the virtual detector generation module, so as to eliminate the interference data by the feature extraction manner, reduce the auditing difficulty of each virtual detector, and improve the auditing accuracy of the work content.
[0064] The virtual detector generation module 102 is configured to generate virtual detectors of multiple detection dimensions based on the received auditing rule text of the running version, and is also configured to call each virtual detector to perform a violation audit on the work content description feature. The virtual detector generation module 102 is an AI agent.
[0065] It should be noted that in an actual application scenario, the AI agent (AI Agent) refers to an artificial intelligence entity that can perceive the environment and take actions to achieve a specific goal. Specifically, the virtual detector generation module 102 can perform semantic recognition on the auditing rule text of the running version, and perform model autonomous training or algorithm combination based on the semantic recognition result, so as to generate virtual detectors of multiple detection dimensions, and then generate virtual detectors adapted to the current auditing requirements based on the auditing rule text of the running version without human intervention, thereby improving the reusability of the content auditing system.
[0066] The application generates a virtual detector by configuring an AI agent as a virtual detector generation module, and generates a virtual detector of multiple detection dimensions based on the received audit rule text of the running version, so that the generated virtual detector can be audited based on the audit rules of the running version, and manual intervention is not required in the generation process of each virtual detector, thereby achieving improved reusability of the work content audit system while ensuring audit accuracy. Moreover, the work description extraction module is configured to extract description features based on the received work, and the extracted work content description features are sent to the virtual detector generation module, so that interference data is eliminated through feature extraction, the audit difficulty of each virtual detector is reduced, and the audit accuracy of the work content is improved. It can be seen that the application improves the reusability and audit accuracy of the work audit system.
[0067] In a possible implementation, the work content audit system provided in the first aspect of the application further includes:
[0068] The semantic search module is configured to perform semantic extraction on the received activity theme text of the running version, and determine a semantic route based on the semantic extraction result.
[0069] It should be noted that in actual application scenarios, the above-mentioned semantic route (Semantic Routing) is an information index address that stores information adapted to the semantic extraction result, which is determined based on the semantic extraction result. Due to the influence of work quality or the accuracy of the work description extraction module, the output accuracy of the extracted work content description features is insufficient. Therefore, the application performs semantic extraction on the received activity theme text of the running version, and determines a semantic route based on the semantic extraction result, so that when information needs to be searched to assist in improving the accuracy of work content description feature generation, the adapted information of the semantic extraction result can be quickly retrieved through the semantic route without manual intervention, thereby improving the information retrieval efficiency.
[0070] It should be noted that in actual application scenarios, the above-mentioned semantic search module can be a virtual module integrated by a large language model and a knowledge graph. The implementation of the above-mentioned semantic search module performing semantic extraction on the received activity theme text of the running version and determining a semantic route based on the semantic extraction result can include the following steps B1 to B3.
[0071] Step B1, store each entity and entity relationship in the historical activity data by using a graph database (such as Neo4j), and construct a knowledge graph based on the graph database. And trigger step B2.
[0072] Step B2, perform semantic extraction on the activity theme text of the running version by using a large language model, and obtain a semantic extraction result. And trigger step B3.
[0073] Step B3, traversing the routing path associated with the semantic extraction result in the knowledge graph by using a query language (such as Cypher) to obtain semantic routing.
[0074] In a possible implementation, the work content review system provided in the first aspect of the present application further includes:
[0075] The task coordination module is configured to identify the information richness of the work before the work description extraction module is used to perform description feature extraction on at least the received work, and send a content supplement instruction to the semantic search module if the information richness is lower than a preset richness threshold.
[0076] It should be noted that in actual application scenarios, the implementation of the task coordination module identifying the information richness of the work can be various, and one example is provided herein.
[0077] Suppose the input work is a dynamic video without audio, and the dynamic video contains image data and text data. The implementation of identifying the information richness of the work includes the following steps C1 to C3.
[0078] Step C1, the task coordination module is used to extract text dimension information and video dimension information of the work respectively, and Step C2 is triggered.
[0079] Those skilled in the art can understand that the text dimension information can include text length, vocabulary richness, semantic complexity, etc. The text dimension information can be achieved by a natural language processing (NLP) algorithm. The video dimension information can include image resolution and color complexity. The text dimension information can be achieved by a grey-level co-occurrence matrix (GLCM) or a generative adversarial network (GAN). The present application does not make too many limitations and repetitions on the specific implementation process of Step C1 and the algorithm type used.
[0080] Step C2, the weights of each sub-dimension in the text dimension information and the video dimension information of the work are weighted and summed to obtain an information richness score.
[0081] It should be noted that in actual application scenarios, the present application configures the task coordination module to identify the information richness of the work before the work description extraction module is used to perform description feature extraction on at least the received work, and sends a content supplement instruction to the semantic search module if the information richness is lower than a preset richness threshold, thereby avoiding the risk of subsequent work content description feature accuracy decline due to insufficient information richness of the work.
[0082] In a possible implementation, the task coordination module can be configured to receive the work, the audit rule text of the running version, and the activity theme text, and forward the above content.
[0083] In a possible implementation, the semantic search module is further configured to search the supplementary description information based on the semantic routing upon receiving the content supplement instruction, and send the supplementary description information to the work description extraction module.
[0084] It should be noted that the present application configures the semantic search module to search the supplementary description information based on the semantic routing and send the supplementary description information to the work description extraction module, so as to realize sufficient collection of content information representing the work, and thus improve the accuracy of the work content description features generated subsequently.
[0085] In a possible implementation, the work description extraction module 101 performs description feature extraction based at least on the received work, including:
[0086] The work description extraction module 101 performs description feature extraction based on the work and the supplementary description information.
[0087] It should be noted that the present application configures the work description extraction module 101 to perform description feature extraction based on the work and the supplementary description information, so as to improve the accuracy of the work content description features generated, and thus improve the subsequent audit accuracy of the illegal content.
[0088] In a possible implementation, to further improve the accuracy of the work content description features, after the work description extraction module 101 performs description feature extraction based on the work and the supplementary description information, the work content description features in the form of text can be displayed, and the work content description features modified by the user operation can be received. Further, the work description extraction module 101 is optimized based on the work content description features modified by the operation, so as to improve the output accuracy of the work description extraction module 101 subsequently.
[0089] In a possible implementation, the work content audit system provided in the first aspect of the present application further includes:
[0090] The illegal content labeling module is configured to receive the illegal audit result of the content sent by the virtual detector, label the work based on the illegal information in the illegal audit result, and output the labeling result.
[0091] It should be noted that in the actual application scenario, the work content labeling module of the present application labels the work based on the violation information in the violation review result, and outputs the labeling result, thereby improving the modification efficiency of the work author on the violation content, shortening the modification waiting time, and further accelerating the overall review efficiency.
[0092] It should be noted that in the actual application scenario, to avoid labeling errors, the labeling result can be sent to the review end for manual review before the labeling result is output, and the labeling result is output to the work author if the review is passed.
[0093] Those skilled in the art can understand that in the actual application scenario, the above-mentioned violation content labeling module can be a virtual module constructed based on a target detection algorithm. The types of the above-mentioned target detection algorithm include but are not limited to YOLO, Faster Region-based Convolutional Neural Networks (FasterR-CN), Support Vector Machine (SVM), etc. The types of the above-mentioned target detection algorithm and the construction process of the above-mentioned violation content labeling module are not limited and described in detail.
[0094] In one possible implementation, the work content review system provided by the first aspect of the present application further comprises:
[0095] The optimization module is configured to send the labeling result to the virtual detector generation module to enable the virtual detector generation module to adjust the parameters of each virtual detector.
[0096] It should be noted that in the actual application scenario, the present application sends the labeling result to the virtual detector generation module through the configuration optimization module, so as to enable the virtual detector generation module to adjust the parameters of each virtual detector, thereby improving the detection accuracy of the virtual detector by increasing the optimization data to optimize the internal parameters of the virtual detector.
[0097] In order to facilitate the understanding of the work content review system of the first aspect of the present application and any possible implementation of the first aspect, the present application is described in combination with one possible implementation of the present application:
[0098] As shown in Figure 3 , it is a signaling diagram of a work content review system configuration stage, as shown in Figure 4 , it is a signaling diagram of a work content review system running stage. The work content review system comprises a task coordination module, a semantic search module, a work description extraction module, a virtual detector generation module, a violation content labeling module and an optimization module.
[0099] AsFigure 3 The specific operation steps of the signaling diagram shown in the configuration phase of a content review system are as follows:
[0100] In step S301, the task coordination module obtains the audit rules text and the activity topic text of the running version. This triggers steps S302 and S303.
[0101] In step S302, the task coordination module sends the activity topic text of the running version to the semantic search module, triggering step S304.
[0102] In step S303, the task coordination module sends the audit rule text of the running version to the virtual detector generation module, triggering step S305.
[0103] In step S304, the semantic search module determines semantic routes based on the semantic extraction results of the active topic text of the running version and stores the semantic routes.
[0104] In step S305, the virtual detector generation module generates virtual detectors for multiple detection dimensions based on the received audit rule text of the running version.
[0105] like Figure 4 The specific operation steps of the signaling diagram shown in the diagram of a content review system during its operation are as follows:
[0106] In step S401, the task coordination module obtains the artwork and triggers step S402.
[0107] In step S402, the task coordination module determines whether the information richness of the work is lower than a preset richness threshold. If yes, step S403 is triggered. If no, step S404 is triggered.
[0108] In step S403, the task coordination module sends a content supplementation instruction to the semantic search module, triggering step S405.
[0109] In step S404, the task coordination module sends the work to the work description extraction module, triggering step S406.
[0110] In step S405, the semantic search module searches for supplementary descriptive information based on semantic routing, thus triggering step S407.
[0111] Step S406: The work description extraction module extracts at least the descriptive features of the work to obtain the content description features of the work. This triggers step S408.
[0112] In step S407, the semantic search module sends supplementary description information to the work description extraction module, triggering step S406.
[0113] Step S408, the work description extraction module sends the work content description features to the virtual detector generation module, and triggers step S409.
[0114] Step S409, the virtual detector generation module calls the virtual detectors to conduct violation review on the work content description features, and triggers step S410.
[0115] In a possible implementation, if the violation review results output by the virtual detectors of each detection dimension in step S409 are all non-violation, the content output is the review result of non-violation.
[0116] Step S410, the virtual detector generation module sends the violation review result to the violation content labeling module, and triggers step S411.
[0117] Step S411, the violation content labeling module labels the work based on the violation review result, and outputs the labeling result, and triggers step S412.
[0118] Step S412, the violation content labeling module sends the labeling result to the optimization module, and triggers step S413.
[0119] Step S413, the optimization module extracts optimization data from the labeling result, and triggers step S414.
[0120] Step S414, the optimization module sends the optimization data to the virtual detector generation module, and triggers step S415.
[0121] Step S415, the virtual detector generation module adjusts the parameters of each virtual detector based on the optimization data.
[0122] The second aspect of the application provides a work content review method, which is applied to the work content review system as described in the first aspect of the application and any possible implementation of the first aspect, and the work content review method comprises the following steps: Figure 5 As shown in the figure, the work content review method comprises the following steps:
[0123] S501, obtaining a work and a review rule text of a running version;
[0124] S502, inputting the review rule text into a virtual detector generation module, so that the virtual detector generation module generates virtual detectors of multiple detection dimensions, and the virtual detector generation module is an AI intelligent agent;
[0125] S503, inputting at least the work into a work description extraction module, so that the work description extraction module extracts description features of the work, and obtains work content description features of the work;
[0126] S504, input the work content description feature into a virtual detector generation module, so that the virtual detector generation module calls each virtual detector to conduct a violation review on the work content description feature.
[0127] In a possible implementation, the work content review system provided in the first aspect of the application further includes a semantic search module, and the work content review method provided in the second aspect of the application further includes:
[0128] The received active topic text of the running version is input into the semantic search module, so that the semantic search module performs semantic extraction on the active topic text of the running version, and determines a semantic route based on the semantic extraction result.
[0129] In a possible implementation, the work content review system provided in the first aspect of the application further includes a task coordination module, and the work content review method provided in the second aspect of the application further includes:
[0130] Before the work description extraction module is used to extract the description feature of at least the received work, the information richness of the work is identified by using the task coordination module, and in a case where the information richness is lower than a preset richness threshold, a content supplement instruction is sent to the semantic search module.
[0131] In a possible implementation, the work content review system provided in the first aspect of the application further includes:
[0132] In a case where the content supplement instruction is received, the semantic search module is called to search for supplement description information based on the semantic route, and the supplement description information is sent to the work description extraction module.
[0133] In a possible implementation, the above at least inputting the work into the work description extraction module so that the work description extraction module extracts the description feature of the work includes:
[0134] The work description extraction module is called to extract the description feature based on the work and the supplement description information.
[0135] In a possible implementation, the work content review system provided in the first aspect of the application further includes a violation content labeling module, and the work content review method provided in the second aspect of the application further includes:
[0136] The violation content labeling module is called to receive the content sent by the virtual detector as a violation review result of existing violations, to label the work based on the violation information in the violation review result, and to output a labeling result.
[0137] In a possible implementation, the work content review system provided in the first aspect of the application further includes an optimization module, and the work content review method provided in the second aspect of the application further includes:
[0138] The optimization module sends the annotation results to the virtual detector generation module so that the virtual detector generation module can adjust the parameters of each virtual detector.
[0139] A third aspect of this application provides an electronic device, including at least one processor and a memory connected to the processor, wherein:
[0140] Memory is used to store computer programs;
[0141] The processor is used to execute computer programs to enable electronic devices to implement the content review method of the second aspect or any implementation thereof described above.
[0142] This application also provides an electronic device in its embodiments. (See reference...) Figure 6 The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 6 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.
[0143] like Figure 6 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. When the electronic device is powered on, the RAM 603 also stores various programs and data required for the operation of the electronic device. The processing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.
[0144] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 608 including, for example, memory cards, hard drives, etc.; and communication devices 609. Communication device 609 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have instead.
[0145] The fourth aspect of the present application provides a computer storage medium, the storage medium carries one or more computer programs, when the one or more computer programs are executed by an electronic device, the electronic device can execute the work content review method of the second aspect or any implementation manner of the second aspect.
[0146] In addition, it should be noted that the above-described apparatus embodiments are merely illustrative, wherein the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. In addition, the connection relationship between the modules in the apparatus embodiments provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines.
[0147] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary general hardware, and of course can also be realized by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. Generally, functions completed by computer programs can be easily realized by corresponding hardware, and the specific hardware structure for realizing the same function can also be various, such as analog circuit, digital circuit or special circuit, etc. However, for the present application, software program implementation is a better embodiment. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., including a plurality of instructions to make a computer device (which can be a personal computer, training device, or network device, etc.) execute the methods described in various embodiments of the present application.
[0148] In the above embodiments, all or part can be realized by software, hardware, firmware or any combination thereof. When realized by software, it can be realized in the form of a computer program product in whole or in part.
[0149] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another, for example, the computer instructions can be transmitted from one website, computer, training device or data center to another website, computer, training device or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be stored by the computer or a data storage device such as a training device, a data center, etc. integrated with one or more available media. The available media can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.
Claims
1. A work content review system characterized by comprising: Comprising: a work description extraction module and a virtual detector generation module; the work description extraction module is configured to perform description feature extraction based on at least the received work, and send the extracted work content description features to the virtual detector generation module; the virtual detector generation module is configured to generate virtual detectors of multiple detection dimensions based on the received audit rule text of the running version, and is further configured to call each virtual detector to perform violation audit on the work content description features, and the virtual detector generation module is an AI agent.
2. The work content review system according to claim 1, characterized in that, The work content audit system further comprises: a semantic search module configured to perform semantic extraction on the received activity theme text of the running version, and determine a semantic route based on the semantic extraction result.
3. The work content review system according to claim 2, characterized in that, The work content audit system further comprises: a task coordination module configured to identify information richness of the work before the work description extraction module performs description feature extraction on at least the received work, and send a content supplement instruction to the semantic search module if the information richness is lower than a preset richness threshold.
4. The work content review system according to claim 3, characterized in that, The semantic search module is further configured to search for supplemental description information based on the semantic route upon receipt of the content supplement instruction, and send the supplemental description information to the work description extraction module.
5. The work content review system according to claim 4, characterized in that, The work description extraction module performs description feature extraction based on at least the received work, comprising: The work description extraction module performs the description feature extraction based on the work and the supplemental description information.
6. The work content review system according to claim 1, characterized in that, The work content audit system further comprises: a violation content labeling module configured to receive a violation audit result sent by the virtual detector that the content exists a violation, label the work based on violation information in the violation audit result, and output a labeling result.
7. The work content review system according to claim 6, characterized in that, The work content audit system further comprises: an optimization module configured to send the labeling result to the virtual detector generation module, so that the virtual detector generation module adjusts parameters of each virtual detector.
8. A work content review method characterized by comprising: The work content audit method applied to the work content audit system of any one of claims 1 to 7, comprising: obtaining a work and an audit rule text of a running version; inputting the audit rule text into a virtual detector generation module, so that the virtual detector generation module generates virtual detectors of multiple detection dimensions, and the virtual detector generation module is an AI agent; inputting at least the work into a work description extraction module, so that the work description extraction module performs description feature extraction on the work to obtain work content description features of the work; inputting the work content description features into the virtual detector generation module, so that the virtual detector generation module calls each virtual detector to perform violation audit on the work content description features.
9. An electronic device, comprising: Comprising at least one processor and a memory connected to the processor, wherein: the memory is configured to store a computer program; the processor is configured to execute the computer program, so that the electronic device can implement the work content audit method of claim 8.
10. A computer storage medium, characterized in that, The storage medium carries one or more computer programs, and when the one or more computer programs are executed by the electronic device, the electronic device can implement the work content review method as claimed in claim 8.