CN-121984892-A - Cross-platform network content intelligent supervision method and system based on multi-mode AI

CN121984892ACN 121984892 ACN121984892 ACN 121984892ACN-121984892-A

Abstract

The invention discloses a multi-mode AI-based cross-platform network content intelligent supervision method and system, which relate to the technical field of network content supervision, realize unified collection and management of cross-platform content by establishing a multi-platform data collection adapter, solve the technical problem that the existing system cannot perform cross-platform supervision, realize deep fusion analysis of texts, images and videos by establishing a multi-mode AI analysis engine, greatly improve the accuracy of content understanding and risk identification, realize accurate risk classification by establishing a hierarchical risk tag system, provide scientific basis for supervision decision, and finally provide comprehensive and visual supervision analysis tools for supervision staff by generating intelligent supervision reports and establishing a real-time monitoring and early warning mechanism, and remarkably improve supervision efficiency and effect. Meanwhile, by adopting ReAct reasoning frames, the system can perform multi-step reasoning analysis, comprehensively judge by combining analysis results of texts, images and videos, avoid the limitation of single-mode analysis and improve the accuracy and reliability of risk identification.

Inventors

LI JIANWEI
CHANG XINYU
LI ZHUO
LING ZHI
ZHANG YIMING
LIU CHUNYANG
SHI JIAN
JIA HONGRU
YAO RUI

Assignees

齐齐哈尔市公安局

Dates

Publication Date: 20260505
Application Date: 20260130

Claims (7)

1. The intelligent cross-platform network content supervision method based on the multi-mode AI is characterized by comprising the following steps: Establishing a multi-platform data acquisition adapter, respectively interfacing with a main stream social platform of trembling, fast handholding and small red books, and acquiring user generated content data comprising text description, pictures and video multimedia information through PLAYWRIGHT automation technology; Establishing a multi-mode AI analysis engine, carrying out semantic analysis on the collected text content by integrating a large language model, carrying out image understanding on pictures and video key frames by integrating a visual model, and identifying risk factors related to yellow, gambling, toxicity, riot, politics, terrorism and gun violation, wherein the multi-mode AI analysis engine adopts ReAct reasoning frames, and outputs comprehensive risk assessment results through the combination judgment of text analysis, image analysis and video analysis; Establishing a hierarchical risk tag system, and classifying the risk grades of the contents according to the AI analysis risk assessment result, wherein the risk grades comprise four grades of high risk, medium risk, low risk and normal contents; Generating an intelligent supervision report, and automatically generating a comprehensive supervision report containing risk content statistics, word cloud analysis, trend analysis and treatment suggestions based on an AI analysis risk assessment result; And a real-time monitoring and early warning mechanism is established, an early warning process is automatically triggered when high-risk content is detected, and early warning information is pushed to supervisory personnel.
2. The method for intelligently supervising cross-platform network content based on multi-modal AI of claim 1, wherein the multi-platform data collection adapter comprises: The tremble data acquisition module is used for acquiring title, description, comment, praise and share data of tremble short videos by simulating user behaviors; The fast hand data acquisition module acquires metadata information of the fast hand short video and user interaction data; The small red book data acquisition module acquires text description, pictures and user interaction information of the picture and text contents of the small red book; And the unified data standardization module is used for converting the data acquired by different platforms into a unified data format and field structure.
3. The multi-modal AI-based cross-platform web content intelligent supervision method as set forth in claim 1, wherein the multi-modal AI analysis engine includes: the text analysis module is used for carrying out semantic understanding and risk element identification on the text content by using the large language model; the image analysis module is used for carrying out target detection and scene understanding on the picture content by using the visual model and identifying illegal visual elements; the video analysis module is used for detecting risk factors of the video content through key frame extraction and image analysis technologies; and the comprehensive research and judgment module is used for carrying out comprehensive risk assessment based on analysis results of the text, the image and the video and outputting final risk grade and classification labels.
4. The multi-modal AI-based cross-platform network content intelligent supervision method as set forth in claim 1, wherein the hierarchical risk tag system comprises: high risk tags, including yellow, gambling, toxic, riot, political, terrorist, gun, serious violation; risk tags, including misstatement, sensitive topics, and potentially risk content; low risk tags, including edge content, content that requires attention; normal tags include health content that meets specifications.
5. The multi-modal AI-based cross-platform network content intelligent supervision method as set forth in claim 1, wherein the intelligent supervision report generation includes: the method comprises the steps of (1) carrying out statistical analysis on risk contents, and counting the quantity, the duty ratio and the distribution condition of various risk contents; word cloud analysis, namely extracting high-frequency keywords and generating a visual word cloud picture; Trend analysis, namely analyzing time distribution and change trend of risk content; treatment suggestions, corresponding treatment suggestions are provided based on the risk level and the content type.
6. The multi-modal AI-based cross-platform network content intelligent supervision method as set forth in claim 1, wherein the real-time monitoring and early warning mechanism comprises: setting risk threshold values, namely setting corresponding early warning threshold values for different risk levels; automatic early warning triggering, wherein the early warning is automatically triggered when high risk content exceeding a threshold value is detected; the early warning information is pushed, and the early warning information is pushed to the supervisory personnel in various modes; and (5) early warning processing tracking, and recording early warning processing processes and results.
7. The cross-platform network content intelligent supervision system based on the multi-mode AI is characterized by comprising a data acquisition layer, an AI analysis layer, a supervision application layer and an early warning push layer; the data acquisition layer is used for establishing a multi-platform data acquisition adapter, respectively butting main stream social platforms of trembling, fast handedness and small red books, and acquiring user generated content data; The AI analysis layer is used for constructing a multi-mode AI analysis engine, intelligently analyzing the collected text, picture and video content and identifying illegal risk elements; The supervision application layer is used for establishing a hierarchical risk tag system, generating an intelligent supervision report and providing supervision data inquiry and statistical analysis functions; the early warning pushing layer is used for establishing a real-time monitoring early warning mechanism, and automatically triggering an early warning process when high risk content is detected.

Description

Cross-platform network content intelligent supervision method and system based on multi-mode AI Technical Field The invention relates to the technical field of network content supervision, in particular to a cross-platform network content intelligent supervision method and system based on multi-mode AI. Background With the rapid development of internet technology and the popularization of social media, network content supervision has become an important task for maintaining network space security and protecting user interests. Currently, mainstream social platforms such as tremble, fast-handed, reddish books and the like generate a huge amount of user-generated content each day, including various forms such as texts, pictures, videos and the like. Such content may include offensive information such as yellow, gambling, toxic, riot, political, etc., which poses a threat to network space security and social stability. Traditional network content supervision mainly relies on manual auditing, and has the problems of low efficiency, high cost, strong subjectivity and the like. With the development of artificial intelligence technology, automatic content auditing technology based on machine learning is gradually rising. However, the existing network content supervision system has the following technical problems that firstly, the existing system mainly carries out supervision on single-platform or single-mode content, lacks cross-platform unified supervision capability, cannot realize unified management and analysis on a plurality of mainstream social platforms, secondly, the existing system mainly lacks deep fusion analysis capability on multi-mode content based on a traditional machine learning method, cannot fully utilize associated information among texts, images and videos to carry out comprehensive judgment, thirdly, the existing system lacks intelligent supervision report generation capability, cannot provide visual and comprehensive supervision analysis reports for supervision staff, and finally, the risk identification precision and efficiency of the existing system are required to be improved, and misjudgment is easy to occur particularly when complex and fuzzy illegal content is processed. On the other hand, with the rapid development of large language models and visual model technologies, the multi-modal AI technology provides a new technological path for web content supervision. The multi-mode AI can process information of multiple modes such as text, image, video and the like at the same time, and improves accuracy of content understanding and analysis through information fusion and complementation among the modes. However, how to effectively apply the multi-mode AI technology to cross-platform network content supervision, how to construct an efficient multi-mode analysis engine, and how to realize intelligent supervision report generation are still technical problems to be solved. Disclosure of Invention The invention aims to provide a cross-platform network content intelligent supervision method and system based on multi-mode AI. According to the invention, cross-platform content unified collection is realized by establishing a multi-platform data collection adapter, a multi-mode AI analysis engine is established to realize deep fusion analysis of texts, images and videos, a hierarchical risk tag system is established to realize accurate risk classification, an intelligent supervision report is generated to provide comprehensive supervision analysis, and a real-time monitoring and early warning mechanism is established to realize timely risk response. Therefore, the automatic intelligent supervision of the cross-platform network content is realized, and the supervision efficiency and accuracy are greatly improved. In order to achieve the above object, the present invention provides the following technical solutions: A multi-mode AI-based cross-platform network content intelligent supervision method comprises the following steps: Establishing a multi-platform data acquisition adapter, respectively interfacing with main stream social platforms such as tremble voice, fast handhold, reddish books and the like, and acquiring user generated content data including multimedia information such as text description, pictures, videos and the like through PLAYWRIGHT automation technology; Establishing a multi-platform data acquisition adapter, respectively interfacing with a main stream social platform of trembling, fast handholding and small red books, and acquiring user generated content data comprising text description, pictures and video multimedia information through PLAYWRIGHT automation technology; Establishing a multi-mode AI analysis engine, carrying out semantic analysis on the collected text content by integrating a large language model, carrying out image understanding on pictures and video key frames by integrating a visual model, and identifying risk factors related to yellow, gambling, toxicity,