By Akos Lada, Data Science Manager, Meihong Wang, Engineering Director, and Tak Yan, hàng hóa Management Director
*

When it comes khổng lồ the News Feed algorithm, there are many theories và myths. Most people understand that there’s an algorithm at work, và many know some of the factors that inform that algorithm (whether you like a post or engage with it, etc.). But there’s still quite a lot that’s misunderstood. 

We publicly nói qua many of the details and features of News Feed. But under the hood, the machine learning (ML) ranking system that powers News Feed is incredibly complex, with many layers. We are sharing new details vuonxavietnam.net how our ranking system works và the challenges of building a system to personalize the nội dung for more than 2 billion people and show each of them content that is relevant & meaningful for them, every time they come khổng lồ Facebook.

Bạn đang xem: Web13_news feed nghĩa là gì, news feed Để làm gì, dành cho những ai chưa biết

What’s So Hard vuonxavietnam.net This?

First, the volume is enormous. More than 2 billion people around the world use Facebook. For each of those people, there are more than a thousand “candidate” posts (or posts that could potentially appear in that person’s feed). We are now talking vuonxavietnam.net trillions of posts across all the people on Facebook. 

Now consider that for each person on Facebook, there are thousands of signals that we need to evaluate to lớn determine what that person might find most relevant. So we have trillions of posts & thousands of signals — và we need khổng lồ predict what each of those people wants to see in their feed instantly. When you xuất hiện up Facebook, that process happens in the background in just the second or so it takes khổng lồ load your News Feed. 

And once we’ve got all this working, things change, & we need khổng lồ factor in new issues that arise, such as clickbait và the spread of misinformation. When this happens, we need khổng lồ find new solutions. In reality, the ranking system is not just one single algorithm; it’s multiple layers of ML models & rankings that we apply in order khổng lồ predict the nội dung that’s most relevant & meaningful for each user. As we move through each stage, the ranking system narrows down those thousands of candidate posts lớn the few hundred that appear in someone’s News Feed at any given time.

*

How Does it Work?


Since Juan’s login yesterday, his friend Wei posted a photo of his cocker spaniel. Another friend, Saanvi, posted a đoạn phim from her morning run. His favorite Page published an interesting article vuonxavietnam.net the best way to lớn view the Milky Way at night, while his favorite cooking Group posted four new sourdough recipes. 

All this nội dung is likely to be relevant or interesting khổng lồ Juan because he has chosen to lớn follow the people or Pages sharing it. Khổng lồ decide which of these things should appear higher in Juan’s News Feed, we need to predict what matters most khổng lồ him & which nội dung carries the highest value for him. In mathematical terms, we need to define an objective function for Juan & perform a single-objective optimization.

We can use the characteristics of a post, such as who is tagged in a photo & when it was posted, khổng lồ predict whether Juan might like it. For example, if Juan tends to lớn interact with Saanvi’s posts (e.g., sharing or commenting) often & her running clip is very recent, there is a high probability that Juan will lượt thích her post. If Juan has engaged with more đoạn clip content than photos in the past, the lượt thích prediction for Wei’s photo of his cocker spaniel might be fairly low. In this case, our ranking algorithm would rank Saanvi’s running clip higher than Wei’s dog photo because it predicts a higher probability that Juan would lượt thích it.

But liking is not the only way people express their preferences on Facebook. Every day, people tóm tắt articles they find interesting, watch videos from people or celebrities they follow, or leave thoughtful comments on their friends’ posts. Mathematically, things get more complex when we need khổng lồ optimize for multiple objectives that all địa chỉ up lớn our primary objective: creating the most long-term value for people by showing them nội dung that is meaningful & relevant to lớn them.

Xem thêm: Hydrogenated Castor Oil, Peg 40 Hydrogenated Castor Oil Là Gì

Multiple ML models produce multiple predictions for Juan: the probability that he’ll engage with Wei’s photo, Saanvi’s video, the Milky Way article, or the sourdough recipes. Each mã sản phẩm tries to rank these pieces of content for Juan. Sometimes they disagree — there might be a higher probability that Juan would lượt thích Saanvi’s running video clip than the Milky Way article, but he might be more likely to comment on the article than on the video. So we need a way lớn combine these varying predictions into one score that’s optimized for our primary objective of long-term value.

How can we measure whether something creates long-term value for a person? We ask them. For example, we survey people lớn ask how meaningful they found an interaction with their friends or whether a post was worth their time so that our system reflects what people say they enjoy and find meaningful. Then we can take each prediction into trương mục for Juan based on the actions that people tell us (via surveys) are more meaningful & worth their time.

Peeling Back the Layers

To rank more than a thousand posts per user, per day, for more than 2 billion people — in real time — we need to lớn make the process efficient. We manage this in various steps, strategically arranged to make it fast and to limit the amount of computing resources required. 

First, the system collects all the candidate posts we can possibly rank for Juan (the cocker spaniel photo, the running video, etc.). This eligible inventory includes any post shared with Juan by a friend, Group, or Page he’s connected to lớn that was made since his last login & has not been deleted. But how should we handle posts created before Juan’s last login that he hasn’t seen yet? 

To make sure unseen posts are reconsidered, we apply an unread bumping logic: Fresh posts that were ranked for Juan (but not seen by him) in his previous sessions are added lớn the eligible inventory for this session. We also apply an action-bumping ngắn gọn xúc tích so that any posts Juan has already seen that have since triggered an interesting conversation among his friends are added lớn the eligible inventory as well. 

Next, the system needs lớn score each post for a variety of factors, such as the type of post, similarity to other items, & how much the post matches what Juan tends khổng lồ interact with. Lớn calculate this for more than 1,000 posts, for each of the billions of users — all in real time — we run these models for all candidate stories in parallel on multiple machines, called predictors.

Before we combine all these predictions into a single score, we need khổng lồ apply some additional rules. We wait until after we have these first predictions so that we can narrow the pool of posts to be ranked — & we apply them over multiple passes to lớn save computational power. 

First, certain integrity processes are applied to every post. These are designed to lớn determine which integrity detection measures, if any, need khổng lồ be applied to lớn the stories selected for ranking. In the next pass, a lightweight mã sản phẩm narrows the pool of candidates to approximately 500 of the most relevant posts for Juan. Ranking fewer stories allows us lớn use more powerful neural network models for the next passes. 

Next is the main scoring pass, where most of the personalization happens. Here, a score for each story is calculated independently, và then all 500 posts are put in order by score. For some, the score may be higher for likes than for commenting, as some people lượt thích to express themselves more through liking than commenting. Any kích hoạt a person rarely engages in (for instance, a lượt thích prediction that’s very close to zero) automatically gets a minimal role in ranking, as the predicted value is very low. 

Finally, we run the contextual pass, in which contextual features like content type diversity rules are added lớn make sure Juan’s News Feed has a good mix of content types and he’s not seeing multiple video posts, one after another. All these ranking steps happen in the time it takes for Juan to open the Facebook app, and within seconds, he has a scored News Feed that’s ready for him to browse and enjoy.

Bài viết liên quan

Trả lời

Email của bạn sẽ không được hiển thị công khai. Các trường bắt buộc được đánh dấu *