Reading special draft | What is the algorithm?Internet factory open algorithm mechanism

On August 12, the National Internet Information Office released the name and filing number of the domestic Internet information service algorithm. The algorithm information used by related enterprises can be queried through the Internet information service algorithm filing system.

Today we will talk about the principles of platform algorithms.

In March of this year, the Internet information service algorithm filing system was launched. According to the requirements, algorithm recommendation service providers with public opinion attributes or social mobilization capabilities shall be filed for three parts of algorithm, algorithm information, product and functional information on the date of ten working days.

The filing list on this time shows that there are currently 30 records covering major Internet platforms, such as Tencent, Baidu, Douyin, Taobao, Meituan, etc. In the inquiry system, netizens can clearly understand information about the algorithm principles and operating mechanisms, application scenarios of each platform.

This time, the algorithm is divided into five major categories: personalized push, retrieval filtering, sorting selection, scheduling decision -making, and generating synthetic categories. Personalized push categories are big heads, with 18 records.

In terms of individual recommendation algorithms, many platforms will emphasize that measures should be taken to avoid excessive concentration of the same type of content and cause information cocoons. For example, the record of the jitter algorithm shows that on the one hand, each recommendation will choose the content category that users do not often watch in the past; on the other hand, in the process of obtaining the recommended content, a random random will be added to each time Content to ensure the diversity of content visible content.

Weibo filing a personalized push and hot search algorithm

In terms of algorithms, most platforms mainly meet the user's personalized needs, improve efficiency, and recommend content that may be interested in. But Tmall's product ranking algorithm emphasizes "objective" — hopes to help customers make purchase decisions through the objective ranking data of the platform.

The algorithm was once new, and once "hugged the pipa half -covered noodles. Takeaway riders, online car drivers and other new Internet formats may have been" trapped "by the algorithm for various reasons. It can be said that it is an important step in compacting the responsibility of the platform and protecting the user's rights. However, on the one hand, users are concerned about the negative impact of the algorithm recommendation, but on the other hand, the technical principle of the algorithm "dark box" may not really pay attention Or really can understand.

In general, the supervision of algorithm governance is still led by the government. The implementation of algorithm governance mainly depends on the platform to implement responsibilities. At the same time, as an expert, I also hope that there are more participation opportunities. There are not many filing. Interested people can check the website of the National Network Information Office. The more people pay attention to the algorithm issues, the more likely it is.

- END -