TikTok ANALYZES HOW THE BACKEND RECOMMENDATION SYSTEM OF IT’S ‘FOR YOU’ FEED FUNCTION
The short music video social network TikTok analyzes how the personalized individual user’s main-stream content, the “For You” feed function. For instance, when a user opens up the short-form video app, the user will be faced with a stream of popular videos, tailored to its interest, while on the other hand, in a situation when a user’s friend launches TikTok on his or her device, they will be faced with something totally different. This, TikTok calls main-stream of content the “For You” feed. But, how the backend recommendation system function has been unknown until this day.
TikTok has finally stepped forward to analyze the factors that contribute to the ‘For You’ feed, considering how they are weighed for every single user. The company also explained on how effort is made to ensure the system doesn’t create filter bubbles, which appears as a homogeneous stream of videos, as the ‘For You’ feed is enabled by user input.
Looking at how TikTok app works, the app function by taking into account the videos a user shared or like, the accounts the user follow, and the comments its user post coupled with the content one create to help figure out the user’s interests. On top of it is the backend recommendation system, which collects video information such as captions, sounds, and hashtags that are associated with the content the user like, which, in a small proportion, includes the account settings information and the user’s device.
The user’s device and account settings information such as country setting, language preference and device type are also notable factors but receive low weight recommendation system compared with core factors on how individual’s like is obtained. Though, there are other contributing factors that help TikTok figure out what a user like, factors viewed as strong or weak signal. For optimum system performance and factors that contribute to the platform’s strong signal based on TikTok analysis, for instance, when a user watches a long video from beginning to end is viewed as strong signal of interest, and would be attached with a greater weight as though, the viewer and poster were from the same country.
There is also another thought that a video could receive more views probably uploaded by an account that has higher followers, is based on the fact that it has a larger base of viewers. While on the other hand, the volume of followers or the previous high-performance videos of a user’s account is considered a direct factor in its recommendation system. Other factors could potentially tailor a user’s TikTok experience based on new areas of interest. This happens in the process of using TikTok, it’s system notes the user’s new area of interests and changes, as far as noting when the user follows a new account, explore sounds, hashtag, effects, in addition to trending topics on its Discover tab.
You can possibly notify TikTok of your explicit likes and dislikes by long pressing, areas they could either add a video to their favorites or have it marked “Not interested”.
Prior to how users could get started with the subject app, a concern TikTok take very serious is the root cause of “filter bubble” that result to an “increasingly homogeneous stream of videos” and this occurs when a user’s taste have received so much attention that leads to development of limited experience.
Meanwhile, in a bid to get rid of the cold-start problem, it detailed (TikTok analyzes) app powered by user input and signal, and how initial recommendations are tailored. However, once TikTok is launched on a user’s device, it does not know the actual content individual likes. To fix this, the app would ask new users to select a category of interest to help tailor initial recommendations. But if the user declines to select categories – TikTok automatically presents a general feed of popular videos until it has amassed appreciable input. And as soon as it acquires the first set of likes, comments, and replays, the app starts to initiate its early stage of recommendations.
Also to allow users stumble upon new content categories including new creators, TikTok automatically add video to a user’s ‘For You’ at times that doesn’t seem useful to a user expressed interests or have acquired a good number of likes. This, the company says will allow users to “experience new perspectives and ideas”.
While in a blog post, the company stated that, “By offering different videos from time to time, the system is also able to get a better sense of what’s popular among a wider range of audiences to help provide other TikTok users a great experience, too”. Adding that, “Our goal is to find a balance between suggesting content that’s relevant to you while also helping you find content and creators that encourage you to explore experiences you might not otherwise see”. – TikTok.
Howbeit, it’s understood the disclosure on how the algorithm works comes at a time when top U.S. tech companies are face with antitrust investigations in the United States and EU – coupled with TikTok specifically under United States congressional review based on ties with China.
What’s your take as TikTok analyzes how backend algorithm works?
Be the first to comment