Please select language
12-12 2022
What is the T-Tok recommendation algorithm In fact, it is ByteDance's recommendation engine, how awesome is this recommendation engine, look at today's T-bar, Tencent raised the resources of the whole company to play today's T-bar, and finally rose strongly. Take a look at ByteDance's official description of the recommendation algorithm To operate content on other platforms and play with traffic, the most important thing is to understand the rules of the platform, don't think about it, just read the official documentation. How the D tone recommendation algorithm is a problem that students who do D sound short video operation are very concerned about, D sound official did not disclose the formal algorithm, but with the wisdom of the people and some of the officially disclosed information, netizens have summarized the secret of the D sound recommendation algorithm. The information compiled here is as follows
Machine audit (detect violations or duplicates): Generally, the artificial intelligence model set up in advance is used to identify your video images and keywords.
It has two main key roles: First, check whether there are violations in the work and copywriting, and if it is suspected, it will be intercepted by the machine, and prompt manual attention through yellow, red, etc.; Second, by extracting the pictures and key frames in the video, matching and eliminating weight with the massive works that already exist in the D sound database, and low-traffic recommendations are made for works with duplicate content, or downgraded recommendations (only fans are visible, only themselves are visible). Human review (detection of titles, covers, and keyframes): mainly concentrated in 3 blocks: 1. Video title
2. Screenshot of the cover
3. Video keyframes.
For machine review to screen out suspected illegal works, as well as works in areas prone to violations, D sound reviewers conduct detailed review one by one. If a violation is determined, penalties such as deleting videos, degrading notices, and banning accounts will be imposed according to the offending account.
Step 2: Cold start The recommendation algorithm mechanism of D tone is a well-known information flow funnel algorithm and the core algorithm of today's T bars. After passing the review, the first step is called cold start traffic pool exposure, for example, if you upload a video today, the work that passes the double review, the system will assign you an initial traffic pool: 200-300 online users (there may also be thousands of exposures). Whether you are large or not, as long as you have the ability to produce quality content, you have the opportunity to compete with the big one. Step 3: Data weighting D tone will analyze whether to weight you based on the data produced by these 1000 exposures and combined with your account score, such as completion rate, likes, followers, and comments , forwards, turn fans, tour depth, etc. All of the above will affect your short video data, and make a judgment on whether to weight your short video, and then select the top 10% of videos and add another 10,000 impressions. Step 4: Step 3: Increase the recommendation This step will give greater weight to short videos with good data, and will strengthen crowd tag distribution in the third step to make content distribution more accurate. This is similar to guessing the marking you like, the video is labeled, the user is also tagged, and the label matching will be done between the two. Step 5: Enter the boutique recommendation pool Enter the boutique recommendation pool, large-scale exposure, once you enter the boutique recommendation, the crowd tag is weakened, just like the gentle video of the year, Almost every D sound user will swipe the gentle video. This article is published by the columnist, and the copyright belongs to the original author. The article is the author's personal opinion, does not represent the position of the scholar, please contact the original author for reprinting. If you have any questions, please contact sales@shusheng.com