4 ms·

Is there any recommendation system people we actually happy with? They all seem to suck in my experience

by endisneigh 5y ago

mrfox321 5y ago
TikTok
- colesantiago 5y ago
  Why TikTok in particular? What is the engineering story behind TikTok's recommendation system? How did they get it right?
  - keewee7 5y ago
    TikTok seem to be learning from what the user is actually watching and for how long and not just the user's "Like"/"Not Interested In" actions. However it still seem to learn from the "Not Interested In" action more than any other platform.
    - pedrosorio 5y ago
      This is a pretty misinformed take when it’s publicly known that YouTube was already doing this (learn from what the user is watching and for how long) the year Bytedance was founded (2012): https://blog.youtube/news-and-events/youtube-now-why-we-focus-on-watch-time/ https://blog.youtube/news-and-events/youtube-now-why-we-focu...
      - monkeybutton 5y ago
        Somehow they're doing it better. At least subjectively, people complain more about the YouTube algo's performance than tiktok. For the latter, the most common complaint is that it's too good.
        leobg 5y ago
        TikToks are shorter. So in any given session, a user will give a lot more feedback to TikTok than to YouTube.
        pedrosorio 5y ago
        Bingo
  - gammarator 5y ago
    This article suggests that the “one video at a time” format makes attention metrics much more reliable: https://www.eugenewei.com/blog/2020/9/18/seeing-like-an-algorithm https://www.eugenewei.com/blog/2020/9/18/seeing-like-an-algo...
  - faizshah 5y ago
    Not a direct answer to your question, but at least in English there isn’t that much stuff on how TikTok’s recommendation system works internally. This is the best breakdown I have found on TikTok’s recommendation system internals: https://leehanchung.github.io/2020-02-18-Tik-Tok-Algorithm/ https://leehanchung.github.io/2020-02-18-Tik-Tok-Algorithm/ There’s a more technical recent paper from bytedance as well: https://arxiv.org/pdf/2007.07203.pdf https://arxiv.org/pdf/2007.07203.pdf And another recent one on bytedance user profile system, this paper gives the deepest understanding of their recommendation system: https://www.cs.princeton.edu/courses/archive/spring21/cos598D/icde_2021_camera_ready.pdf https://www.cs.princeton.edu/courses/archive/spring21/cos598... There’s a good breakdown of the recent nytimes article on TikTok’s internals here: https://read.deeplearning.ai/the-batch/issue-122/ https://read.deeplearning.ai/the-batch/issue-122/ If anyone has found anything better let me know. If you try searching yourself you will want to try switching between the keywords “douyin” “toutiao” “bytedance” and “tiktok.”
chudi 5y ago
all feeds are recommendations systems, instagram, facebook, twitter, tiktok, youtube, every single one is a recommendation system.
- notriddle 5y ago
  Technically, yes, but when they're talking about this sort of thing, they mean "personal recommendation system" or "content-based recommendation system." For example, the HN front page is a recommendation system if you literally mean system-that-recommends-web-pages-to-look-at. But it's not personalized; every visitor sees the same front page. This fundamentally makes it a different sort of thing.
  - chudi 5y ago
    If You have 10000 posts that You have to sort it in some way and the user just going to see 20 of those, the sorting is the recommendation system, people are just used to think of products, movies and songs, but in those platforms the users are the products
    - notriddle 5y ago
      This hardly seems like a reasonable way to characterize Netflix, which has a personal recommendation system, especially compared to HN, which is ad supported yet gives the same recommendations to everyone.
      - samhw 5y ago
        My reading of their comments is that they are trying to say that social media and news media can be characterised as having recommendation systems too, not just song and movie platforms (I don't know who exactly they're arguing against – I've never heard anyone say that recommendation systems can only be for songs and movies). I don't think they're really paying much attention to the dimension you're splitting it along, i.e. whether the recommendations are personalised for each user. The huge important idea they have in their head is that recommendations can apply to user-generated social media content too.
charcircuit 5y ago
YouTube, TikTok, and Twitter all work well for me.
- q-base 5y ago
  Does YouTube actually work for someone? I bit the bullet after not really giving any other input than what I watched and started marking "Not interested" or "Don't recommend channel" but I cannot see that it has had any effect at all it keeps recommending rubbish "sensationalist" videos. I know that they probably optimize for ads etc. but if they actually showed me videos I would like to see, then I would spend more time on the platform.
  - charcircuit 5y ago
    Yes, it works great for me. I don't really use the negative feedback. Most recommendation I guess are new uploads from creators I like or videos I haven't watched from creators I like. It shows me videos from unknown creators to me that match my interests as they go viral. It shows some random old videos that just randomly go viral again. It's a good experience.
    - q-base 5y ago
      That's interesting and perhaps I should not be that surprised. If it did not work at all, then they would probably have changed it. Their KPI is probably ad revenue, but still if it worked for no one, then I would expect their revenue to decline as people spend their time elsewhere. But then again, where will you go and watch medium to long videos created by "normal" people? Vimeo? Perhaps youtube's interactions are despite their algorithm and not because.
      - charcircuit 5y ago
        >ad revenue Not directly, most people believe they optimize for session time. It tries to serve you a result that will keep you on the platform watching videos as opposed to needing to keep scrolling or leaving the platform. They truely do want to serve the best videos that they can to you that they think you will be interested in watching. Thankfully for YouTube you watching videos and YouTube getting ad rev is correlated. >where will you go and watch medium to long videos created by "normal" people? No one forces you to watch that format of video. You can go on TikTok, Twitch, Netflix, etc and be entertained.

Is there any recommendation system people we actually happy with? They all seem to suck in my experience

mrfox321 5y ago

TikTok

colesantiago 5y ago
Why TikTok in particular? What is the engineering story behind TikTok's recommendation system? How did they get it right?
- keewee7 5y ago
  TikTok seem to be learning from what the user is actually watching and for how long and not just the user's "Like"/"Not Interested In" actions. However it still seem to learn from the "Not Interested In" action more than any other platform.
  - pedrosorio 5y ago
    This is a pretty misinformed take when it’s publicly known that YouTube was already doing this (learn from what the user is watching and for how long) the year Bytedance was founded (2012): https://blog.youtube/news-and-events/youtube-now-why-we-focus-on-watch-time/ https://blog.youtube/news-and-events/youtube-now-why-we-focu...
    - monkeybutton 5y ago
      Somehow they're doing it better. At least subjectively, people complain more about the YouTube algo's performance than tiktok. For the latter, the most common complaint is that it's too good.
      - leobg 5y ago
        TikToks are shorter. So in any given session, a user will give a lot more feedback to TikTok than to YouTube.
        pedrosorio 5y ago
        Bingo
- gammarator 5y ago
  This article suggests that the “one video at a time” format makes attention metrics much more reliable: https://www.eugenewei.com/blog/2020/9/18/seeing-like-an-algorithm https://www.eugenewei.com/blog/2020/9/18/seeing-like-an-algo...
- faizshah 5y ago
  Not a direct answer to your question, but at least in English there isn’t that much stuff on how TikTok’s recommendation system works internally. This is the best breakdown I have found on TikTok’s recommendation system internals: https://leehanchung.github.io/2020-02-18-Tik-Tok-Algorithm/ https://leehanchung.github.io/2020-02-18-Tik-Tok-Algorithm/ There’s a more technical recent paper from bytedance as well: https://arxiv.org/pdf/2007.07203.pdf https://arxiv.org/pdf/2007.07203.pdf And another recent one on bytedance user profile system, this paper gives the deepest understanding of their recommendation system: https://www.cs.princeton.edu/courses/archive/spring21/cos598D/icde_2021_camera_ready.pdf https://www.cs.princeton.edu/courses/archive/spring21/cos598... There’s a good breakdown of the recent nytimes article on TikTok’s internals here: https://read.deeplearning.ai/the-batch/issue-122/ https://read.deeplearning.ai/the-batch/issue-122/ If anyone has found anything better let me know. If you try searching yourself you will want to try switching between the keywords “douyin” “toutiao” “bytedance” and “tiktok.”

chudi 5y ago

all feeds are recommendations systems, instagram, facebook, twitter, tiktok, youtube, every single one is a recommendation system.

notriddle 5y ago
Technically, yes, but when they're talking about this sort of thing, they mean "personal recommendation system" or "content-based recommendation system." For example, the HN front page is a recommendation system if you literally mean system-that-recommends-web-pages-to-look-at. But it's not personalized; every visitor sees the same front page. This fundamentally makes it a different sort of thing.
- chudi 5y ago
  If You have 10000 posts that You have to sort it in some way and the user just going to see 20 of those, the sorting is the recommendation system, people are just used to think of products, movies and songs, but in those platforms the users are the products
  - notriddle 5y ago
    This hardly seems like a reasonable way to characterize Netflix, which has a personal recommendation system, especially compared to HN, which is ad supported yet gives the same recommendations to everyone.
    - samhw 5y ago
      My reading of their comments is that they are trying to say that social media and news media can be characterised as having recommendation systems too, not just song and movie platforms (I don't know who exactly they're arguing against – I've never heard anyone say that recommendation systems can only be for songs and movies). I don't think they're really paying much attention to the dimension you're splitting it along, i.e. whether the recommendations are personalised for each user. The huge important idea they have in their head is that recommendations can apply to user-generated social media content too.

charcircuit 5y ago

YouTube, TikTok, and Twitter all work well for me.

q-base 5y ago
Does YouTube actually work for someone? I bit the bullet after not really giving any other input than what I watched and started marking "Not interested" or "Don't recommend channel" but I cannot see that it has had any effect at all it keeps recommending rubbish "sensationalist" videos. I know that they probably optimize for ads etc. but if they actually showed me videos I would like to see, then I would spend more time on the platform.
- charcircuit 5y ago
  Yes, it works great for me. I don't really use the negative feedback. Most recommendation I guess are new uploads from creators I like or videos I haven't watched from creators I like. It shows me videos from unknown creators to me that match my interests as they go viral. It shows some random old videos that just randomly go viral again. It's a good experience.
  - q-base 5y ago
    That's interesting and perhaps I should not be that surprised. If it did not work at all, then they would probably have changed it. Their KPI is probably ad revenue, but still if it worked for no one, then I would expect their revenue to decline as people spend their time elsewhere. But then again, where will you go and watch medium to long videos created by "normal" people? Vimeo? Perhaps youtube's interactions are despite their algorithm and not because.
    - charcircuit 5y ago
      >ad revenue Not directly, most people believe they optimize for session time. It tries to serve you a result that will keep you on the platform watching videos as opposed to needing to keep scrolling or leaving the platform. They truely do want to serve the best videos that they can to you that they think you will be interested in watching. Thankfully for YouTube you watching videos and YouTube getting ad rev is correlated. >where will you go and watch medium to long videos created by "normal" people? No one forces you to watch that format of video. You can go on TikTok, Twitch, Netflix, etc and be entertained.