Welcome to visit NetEase Cloud Community to learn more about NetEase’s technical product operation experience.


Manual auditing is important for both station A and station B.


AcFun Standard for Manuscript Review (V1.1) indicates that there are A lot of things to review, in addition to the compliance and tonality of the content itself, there are also picture and sound quality, crash problems, video information problems, etc., and station A does use manual review.


Reference of site B review: From the IPO of Site B, IQiyi and Inke, we can see the current situation of content review of domestic video companies



Say again the general idea that video website examines content:


The content audit of video website mainly has two purposes, one is compliance, the other is copyright. Among them, compliance issues cover a wider range, including pornography, violence and terrorism, and political involvement. Of course, content should also be in line with the needs of business value, or in line with the site positioning, such as advertising, abuse, etc., which has a great impact on user experience and business realization. Video websites usually complete the audit work through the combination of technical means and manual audit. Whether it is filtering in advance or marking some content to let the reviewers focus on, the audit system is indispensable. All non-live content is reviewed before it is released. In addition, reporting channels and fast processing are also necessary.


Video review is a difficult point, but video websites need to review not only videos, but also user profile pictures, nicknames, signatures, titles, profiles, covers, labels, comments, bullets, messages and other images/text content that users can input by themselves. The work is so complex that the audit system needs to be comprehensive, strategic and progressive to ensure the efficiency and accuracy of the audit.


  • Omnipotent means using various algorithms to deal with different content media such as text, images, audio and video, as well as different types of violations such as pornography and violence. In this regard, advances in artificial intelligence provide a powerful weapon. Reference: How does artificial intelligence recognize a yellow map?

  • The strategy not only uses regular expressions and intelligent algorithms to deal with the content itself, but also identifies the risk level of users by IP, terminal, email, mobile phone number and other information, and then makes different treatments, such as limiting the publishing interval. Note that AI can be relied upon, not relied upon.

  • The improvement is due to the fact that the black Ash Industry tries to test and break through the existing system’s defenses. For example, there have been some cases of small-language, pseudo-low-level attacks that bypass YouTube’s audit. Operators and algorithm engineers need to work together to continuously improve the system’s sensitive theshours, logic and algorithms.


In a word, the mature audit system, mature audit team, the coverage of application scenarios, are trained through hundreds of battles.


NetEase Cloud E-Shield intelligent anti-garbage service, based on NetEase’s 20 years of professional anti-garbage experience, as well as the accumulated massive data of 100 million levels, provides customers with extremely fast and intelligent cloud services of advertising filtering, intelligent pornography detection, identification of violence and terrorism, and political detection, helping to build a pure Internet content ecology. Welcome a free trial.


【 recommended 】 “NetEase Big data Lecture Stage 1: Shu Shuo” live event materials: course playback, watching and PPT download