4 ms·

What computer algorithm can be used to code up "...front page position should be determined by quality of conversation not voting"? Are we talking about sentime

What computer algorithm can be used to code up "...front page position should be determined by quality of conversation not voting"? Are we talking about sentiment analysis?

dragonwriter 8y ago
Arguably, the HN flamewar detector is a crude and limited algorithmic approach to exactly that problem. It's obviously not a complete solution, as it seeks neither to address all quality dimensions nor to replace voting, instead being a factor alongside voting, but still...
- jcriddle4 8y ago
  Do you have a link to how the HN flamewar detector works?
  - dragonwriter 8y ago
    Not guaranteed to be current, but the “Controversy” section addresses it: http://www.righto.com/2013/11/how-hacker-news-ranking-really-works.html?m=1 http://www.righto.com/2013/11/how-hacker-news-ranking-really...
    - jcriddle4 8y ago
      Thanks for the excellent link. Quite a good read. From the article: [In order to prevent flamewars on Hacker News, articles with "too many" comments will get heavily penalized as "controversial"....] Ok, that isn't exactly even a slightly sophisticated method that could be extended to decide some comments are "high quality". In theory if you wanted to implement article ranking you need a method that ranks articles based on comment quality. Without a automated method you are really just using humans to do the ranking.
beaconstudios 8y ago
perhaps it could be devolved to the userbase - every user gets one vote per day that can be used to flag a conversation as interesting. If they flag a second discussion, the flag migrates to it (as if to say "actually, this is more interesting"). Display the top N articles in order of number of votes.
- badbug 8y ago
  This assumes the user is accurate in his interpretation of a comment. Once they learn that "a" makes comments more visible and "b" makes them less, it just becomes an agree / disagree vote
  - beaconstudios 8y ago
    I'm not suggesting an upvote-downvote system though, but a per-user "flag" that can highlight 1 article per day. No downvote option, and the UI would highlight its intended use.
    - badbug 8y ago
      The upvote is an implicit down vote for other things. I think you need two things. First an upvote/downvote so that users can express their agree/disagree. This value wouldn't necessarily effect the visibility of the content. Second, a sort of budget system where users can trade earned karma to classify an items "quality". I think this would lead to a system where you would spend your karma on the communities you care/know/like the most. If I were, for example, a blind apple hater, the upvote/dowmvote makes it easy to downvote apple content or upvotecother content to push it down the page. But I may be more inclined to save my karma for promting Android content instead.
badbug 8y ago
"sentiment analysis" is flawed because it lacks broader context of the conversation. A negative/angry/abusive comment can still be a quality comment depending on the comments upstream.
- jcriddle4 8y ago
  Leaving off the problem of negative/angry/.... comments, is the current state of sentiment analysis or AI sophisticated enough to determine that a set of comments are higher quality?
  - badbug 8y ago
    Think about contentious issues like abortion, politics, vim vs emacs and the comments that could arise "Those other people should die in a fire" "But you should shoot them too, just in case" These might be funny, light-hearted, sarcastic, and quality in some topics and angry, mean, toxic in others. I'm sure it is being worked on but I don't see it happening anytime soon. It seems like it would require to much training and that the biases of the trained data would still produce incorrect outcomes.

What computer algorithm can be used to code up "...front page position should be determined by quality of conversation not voting"? Are we talking about sentiment analysis?

dragonwriter 8y ago

Arguably, the HN flamewar detector is a crude and limited algorithmic approach to exactly that problem. It's obviously not a complete solution, as it seeks neither to address all quality dimensions nor to replace voting, instead being a factor alongside voting, but still...

jcriddle4 8y ago
Do you have a link to how the HN flamewar detector works?
- dragonwriter 8y ago
  Not guaranteed to be current, but the “Controversy” section addresses it: http://www.righto.com/2013/11/how-hacker-news-ranking-really-works.html?m=1 http://www.righto.com/2013/11/how-hacker-news-ranking-really...
  - jcriddle4 8y ago
    Thanks for the excellent link. Quite a good read. From the article: [In order to prevent flamewars on Hacker News, articles with "too many" comments will get heavily penalized as "controversial"....] Ok, that isn't exactly even a slightly sophisticated method that could be extended to decide some comments are "high quality". In theory if you wanted to implement article ranking you need a method that ranks articles based on comment quality. Without a automated method you are really just using humans to do the ranking.

beaconstudios 8y ago

perhaps it could be devolved to the userbase - every user gets one vote per day that can be used to flag a conversation as interesting. If they flag a second discussion, the flag migrates to it (as if to say "actually, this is more interesting"). Display the top N articles in order of number of votes.

badbug 8y ago
This assumes the user is accurate in his interpretation of a comment. Once they learn that "a" makes comments more visible and "b" makes them less, it just becomes an agree / disagree vote
- beaconstudios 8y ago
  I'm not suggesting an upvote-downvote system though, but a per-user "flag" that can highlight 1 article per day. No downvote option, and the UI would highlight its intended use.
  - badbug 8y ago
    The upvote is an implicit down vote for other things. I think you need two things. First an upvote/downvote so that users can express their agree/disagree. This value wouldn't necessarily effect the visibility of the content. Second, a sort of budget system where users can trade earned karma to classify an items "quality". I think this would lead to a system where you would spend your karma on the communities you care/know/like the most. If I were, for example, a blind apple hater, the upvote/dowmvote makes it easy to downvote apple content or upvotecother content to push it down the page. But I may be more inclined to save my karma for promting Android content instead.

badbug 8y ago

"sentiment analysis" is flawed because it lacks broader context of the conversation. A negative/angry/abusive comment can still be a quality comment depending on the comments upstream.

jcriddle4 8y ago
Leaving off the problem of negative/angry/.... comments, is the current state of sentiment analysis or AI sophisticated enough to determine that a set of comments are higher quality?
- badbug 8y ago
  Think about contentious issues like abortion, politics, vim vs emacs and the comments that could arise "Those other people should die in a fire" "But you should shoot them too, just in case" These might be funny, light-hearted, sarcastic, and quality in some topics and angry, mean, toxic in others. I'm sure it is being worked on but I don't see it happening anytime soon. It seems like it would require to much training and that the biases of the trained data would still produce incorrect outcomes.