Welcome to visit netease Cloud Community to learn more about Netease’s technical product operation experience.
Recently, “accident” platform many, a fish also quietly during the National Day holiday off the rectification. Platform “electric shock”, pornography as one of the main culprits, guilty. Today’s major platforms, websites can be described as “pornographic”.
So how to effectively prevent pornography? Fundamentally speaking, it needs the host to have considerable self-consciousness, and it will take a long time to solve the problem fundamentally. However, we can still restrict the platform rules, intelligent content recognition, account processing and other aspects to prevent the phenomenon of live pornography.
1. Restrictions of live broadcast platform rules
At present, the measures taken by live broadcasting platforms to prevent illegal content have been relatively perfect. In the rules of the platform, the definition and management of anchors’ violations have been detailed and improved. Take Huya Live as an example, pornography related content is defined in more detail:
2.2.5 It is strictly forbidden to perform in any form any content that is pornographic, arouses sexual desire or of low interest, such as:
(1) Display or implicit display of sensitive parts of the body with sexual characteristics in various ways, such as:
A. Deliberately focus the camera on sexually sensitive areas covered by clothing
B. Touching, pulling and other movements with sexual hints;
C. Use sex objects (or other objects easily associated with sex objects) as performance props;
D. Other vulgar and sexually suggestive behaviors
(2) Live broadcast with vulgar, revealing and sexually suggestive makeup, such as:
A. male shirtless;
B. Women wear overly revealing clothing, such as exposing sensitive parts or objects such as breasts and buttocks;
C. Wear sexy uniforms, sexy underwear, revealing clothing, see-through clothing, flesh-colored tights, underwear over the top, half top, fishnet socks, suspenders, triangle shorts or triangle shorts, low-waisted short shorts, clothing with sexually suggestive words or pictures, etc.;
D. It is strictly prohibited to wear the working uniforms of personnel of state organs and military of the People’s Republic of China (including Hong Kong, Macao and Taiwan) for entertainment broadcast;
E. Displaying or exposing tattoos;
F. Other offensive makeup.
(3) perform dances with or likely to cause a vulgar atmosphere, such as:
A. Sticking your butt out for the camera and showing your crotch;
B. Pick up clothes and skirts;
C. rubbing chest, buttock, sensual licking lips;
D. Simulate sexual movements;
E. Draw circles with hips;
F. Focus on sensitive areas.
(4) Live broadcast in places prone to vulgar atmosphere, such as bath centers and nightclubs;
(5) the imitation of pornographic and provocative sounds that arouse others’ sexual desire;
(6) Other vulgar and marginal content.
The platform itself has carried out a series of restrictions on pornographic content in accordance with the legal provisions and the development requirements of its own platform, so as to regulate the live broadcast content of anchors.
2. Real-time identification of pornographic content
At present, most live broadcast platforms are equipped with a content audit mechanism, equipped with a large number of audit supervision, real-time inspection of the content of each live broadcast room, illegal or non-conforming content will be directly dealt with.
The advantage of artificial audit is high accuracy, accurate scale grasp, but the shortcomings are also obvious, first of all, the rapid expansion of the number of direct broadcast room, need to be equipped with more artificial audit team, artificial audit cost is high; Secondly, due to the low timeliness of manual audit, it is difficult to achieve real-time monitoring, and there is a risk that pornographic images that flash by will be missed.
To cope with this situation, live broadcast platforms generally use artificial intelligence technology to detect live broadcast content in real time by combining manual and machine detection methods. Cooperate with manual audit to further control the accuracy of scale and machine testing.
All in all, there are still difficulties in eradicating live pornography, but with the optimization of artificial intelligence technology, we believe there is still a bright future
All the year round stationed in the front line of content security, Yi Dun has rich experience in pornography monitoring, text detection, image detection, audio detection, video detection and other non-dead Angle screening pornographic information ability, has been highly recognized by many platforms, websites. Feel the service effect by yourself. Click the link above for a free trial.
A copy of the ECMAScript 2015 code specification