Who produces garbage on the Internet?

  More than 60% of Chinese people consume content on online platforms, and some of them also play the role of producers.
  With the emergence of Twitter and Weibo, UGC (User Generated Content) has become an important information for netizens to browse on the Internet. According to a report released by the China Online Audiovisual Program Service Association, as of December 2020, the number of online audiovisual users in my country reached 944 million, of which 873 million were short video users.
  The UGC platform has become more popular and universal with the expansion of the mobile Internet, and its side effects also follow – first is the occupation of attention, followed by the rapid disappearance of time and amplified emotions. And among the massive amount of content, being able to find out what can really be called “useful information” is tantamount to panning for gold in the sand.
  Many users have had this experience. Douyin often sees different bloggers reading the exact same copy, preaching or popularizing to the audience, but in fact they don’t know what they are talking about. These are the number-making teams manipulating mass production behind the scenes.
  The production of spam has a mature assembly line, and its confrontation and connivance have also become the norm in the Internet world.
1. Pipeline

  Some producers do not deny that the content they produce is “garbage”.
  After 2018, Douyin Kuaishou has risen rapidly, and information feed advertising has become an extremely popular delivery method. Yu Meng works in an education company, and her job is to make advertisements suitable for “mixing” in Douyin. These advertisements will be placed on the terminals of “target users”, attracting them to click, and then divert traffic to their own educational course APP.
  This job is certainly not creative.
  ”The first is to analyze the data. From the launch of the advertisement to the user’s contact, there are a series of rigid indicators to measure.” The KPIs tied to the dream include but are not limited to click-through rate, conversion rate, 3-second playback, 5-second playback and other data. . The particularity of the education industry makes her feel that her work is somewhat meaningful, but she still feels that these videos are “99% garbage”.
  Doing information flow advertising, without volume, everything is empty talk. In Yumeng’s team, one person can make about 150 videos a month. This means that each person needs to produce 7 pieces of content every day. Each video has its own place – they correspond to different Douyin IDs, and an ID is an advertisement “spot”.
  After uploading a video, Yu Meng will set the corresponding “points” in the background. Geographical location, age, type of mobile phone and even occupation can all be the distinguishing points of “points”.
  The company’s business growth has certainly benefited, but there is no definite reason for where those highly-viewed videos work.
  Yu Meng said that the success of popular short video advertisements has an element of luck, and production and delivery are only basic actions. Under the situation that the homogenization competition of video content is getting stronger and stronger, the follow-up effect still needs to be tested by the market. All in all, quantity is useful. Yu Meng and her colleagues can only produce content over and over again.
  The construction of another part of the pipeline stems from Party A’s demand for “flow”. Comments, likes, retweets – all quantifiable indicators will become “data”, and no one cares about where the 100-word content comes from under the data.

  ”Hot search” is the most important part of Li Hui’s work. When the customer needs it, Li Hui is responsible for finding the navy and the big V to post. From amateur accounts to KOLs, all content releases are clearly priced.
  The handling of negative public opinion is the most typical example. Within an hour or two, Li Hui and his team were able to complete dozens of content templates, and then publish them through the “navy supplier”.
  Making reading volume and popularity is the most basic operation, which belongs to the “data maintenance” business of suppliers. Among all the “water army”, the amateur account is the most, “just click on a comment on a popular Weibo”; if the budget is sufficient, you can buy more advanced zombie fans, and their account weights are higher, Comments will come first.
  Li Hui knew better than anyone what was fake. He is very familiar with the “routines” of the Navy’s copywriting, and when he clicks on the hot search on Weibo, Li Hui can basically judge which entries are bought.
2. Fight against “garbage”

  More “garbage” hiding in the dark.

  They are like worms in the content square: once the savage grows, the entire square will corrode. Search “Zhihu batch comments” and “Xiaohongshu notes on behalf of others” in the QQ group. The organizer uses a low price of 2 yuan to obtain a piece of junk information, and the black and gray industry chain extends.
  Fighting against garbage is Wei Ning’s daily job. As the product manager of a domestic security service company, he and his team provide content security services for many well-known Internet companies. Content related to pornography, politics, and violence and terrorism all fall within the scope of their monitoring.
  Algorithms and models are the key to that. The first to join the fray were the staff in charge of manual tagging, who were familiar with which sites had “spam” material, which was where they collected their samples. There are usually only 1-2 spam features in a sample that can be marked. The marked samples enter the model side, and then carry out subsequent machine learning and training.
  In the past year or two, the junk content in black and gray production has been the focus of monitoring by content security products, such as killing pigs. The people who post this content tend to be from the same gang. Wei Ning said that after detecting the publisher’s device, they will define the account logged on the device as a violation, and then directly ban the device. In this way, the publisher of spam content needs to change accounts or even change mobile phones for the next release, and the cost will be greatly increased.
  ”These publishers often have some common behavioral characteristics, such as the behavior of registering and logging in, the avatar they use, their description and chat content.” To sum up, Wei Ning believes that the best way is to combine business security and content security. Combined, increase the cost of publishing spam.
  Large companies gradually improve their own content review platforms, while many medium-sized content platforms choose to cooperate with Wei Ning’s team. The technical ideas of the audit are mostly similar, but the large factories are equipped with more auditors, and the entire inspection process is completed by “manual + AI”.

  But at the same time, garbage producers will also learn to “progress” and try their best to bypass platform supervision.
  From a platform perspective, production and cleanup are also a process of continuous confrontation.
  A product director who has been engaged in content security services since 2014 told Jiemian News that the manual review rate of the products they developed was within 3%. If a platform has 10,000 content data a day, content auditors only need to focus on reviewing and processing more than 300 of them, which is nearly 30 times more efficient than manual work.
  But algorithms also have their drawbacks. Because the platform relies too much on algorithms, content producers often think about how to “step on” the recommendation points of the algorithm, and the priority is even higher than pleasing users. In the opinion of an operator who has worked at Toutiao, this will also lead to the deformity of the content. For example, some article titles are piled with keywords that are easy to get traffic recommendations, but it is difficult to even smooth the text.
  This is also why Internet companies need a large number of “operational” positions – people must fight against garbage with technology, and there is also a need to “fight with algorithms”.
3. Indulge

  Does “garbage” have to be unnecessary? For large companies, it may not be possible to give a positive answer.
  The experience of interning in a leading mobile phone manufacturer subverted Jiang Lingling’s cognition of the position of “content operation”. Her job responsibilities are to promote the company’s new products. To further refine the work tasks, it is to visit Douban and Tieba every day to find suitable content to “carry” to the new APP.
  The new products are aimed at sinking users. Only content that is “sinking” is popular. The more “shattering the three views”, the more reading stories, such as short stories of parents, marriage and love, and beauty gossip. After finding these contents, Jiang Lingling needs to quickly rewrite them, and publish them in batches in the APP developed by the company, “nonstop irrigation”.
  Liu Yan, who has worked in an entertainment Internet company, has a similar experience. At the beginning of the launch of the music APP developed by the company, Liu Yan’s job was to search for comments in major music playback software, and “match” 5-6 manual comments for each song by moving and rewriting. Before leaving get off work each day, more than 100 songs and comments are aggregated in an excel sheet showing the results of her work.
  Even past the early stages of development, the platform’s attitude towards spam is not black and white. Here, defining “value” requires another set of metrics: the greatest value is the content itself that attracts users without breaking the rules.
  For those repetitive and nutritious junk information, the operation of the above-mentioned Toutiao can feel a certain tolerance of the platform. As an information aggregation platform, Toutiao lacks native hotspots, so it needs a lot of content to support the operation of the entire platform. When the amount of information is large enough, users can search for what they want, and the accuracy of the algorithm can be verified and improved. A business leader once told them in a meeting that operations work is like “fishing in the ocean.”
  Facts have also proved the role of this “inclusiveness”. Public information shows that one year after Toutiao went online, the number of users exceeded 50 million; it has accumulated 600 million activated users within four years of its launch; in Q1 2021, Trustmobile data shows that Toutiao has more than 280 million monthly active users.
  Some hesitant to create the sea also disappeared. A few years ago, a news information product had a sense of crisis under the rapid attack of Toutiao. A person involved in the operation of the product told Jiemian News that she clearly felt that there was no consensus within the team on whether to “tolerate” spam. Long after the APP was launched, content editors were still discussing at various conferences, “Should we send spam content to people in the sinking market?” Today, the APP has been shut down.
  In the face of user growth and high-quality content, the platform side has always maintained a delicate gaming attitude. But as competition and regulation intensify, platforms also realize that a product that wants to thrive in the long run must provide cleaner oceans.
  According to Tech Planet, this year Toutiao has established two new business teams, one of which will be a high-end information platform and intelligent reading tool “Knowing Area”. Kuaishou, which used “earth-flavored video” as its growth engine in the early days, also frequently developed knowledge videos and live broadcast content this year. An entrepreneur decided to devote himself to the incubation of knowledge content this year because “the platform will not feed users garbage forever.”
  However, the nature of commercial companies is to pursue profits, and the nature of human beings is to seek effortless fun and excitement. Outside of technology, the fight against spam is more like an eternal war between humanity and humanity.