6 ms·

Comments like this are why I don't bother getting my tech & science news from anywhere but HN.

mkbnnh 9y ago
Eternal Semptember tho. Give it time.
bduerst 9y ago
I've seen an increase in the number of NLP bots that give snippets, but they're far from consistently being accurate. Are there any services that can give you summations like this, kind of like the WSJ's What's News section?
- whimsy 9y ago
  Coverage isn't great, but these two have managed to get in to my bookmarks: http://thecontext.net/ http://thecontext.net/ https://legiblenews.com/ https://legiblenews.com/ I seem to recall a similar weekly service for HN but I can't recall the name of it.
  - j_s 9y ago
    http://n-gate.com/ http://n-gate.com/ - just kidding! Maybe http://www.hackernewsletter.com/ http://www.hackernewsletter.com/ but that's an email.
    - vog 9y ago
      From the n-gate page: > Pasting n-gate content or generally talking about n-gate on "Hacker" "News" is a violation of the Prime Directive. In other words, they want to publish their rants, but are unwilling to discuss their views with anyone, not even the very people they are criticizing. This may be a valid (although insane) approach for a blog, but somewhat contradicts their stated goal: > The purpose of n-gate is to be a remote island of contemptuous sanity in a sea of ridiculous webshit.
    - kronos29296 9y ago
      Is ngate some kind of let's rain shit on hacker news kind of site? Because it looked like that.
      - vog 9y ago
        This description is unfair and not correct. They also rant about FOSDEM.
- rangibaby 9y ago
  Autotldr on Reddit does a very good job
  - lugg 9y ago
    It's not perfect but does the trick 9 out of 10 times, it just uses tfidf irrc.
- wtallis 9y ago
  Sometimes, the best summary is "this article is bunk". You won't get that from a bot. The fatal flaws in bad journalism are too often subtle and require careful parsing or outside knowledge to identify.
  - pas 9y ago
    I think it's easily quantifiable what's a good article. Clear statement at the top that refers to big things, not a storytelling bullshit opening, then the body of the article provides support for that top statement, and later portions are basically just more of the same gradually unfolding sub-articles. If this structure is lacking, it'll be obvious to sentiment analysis, that there are too much free floating statement, and the article either requires too much outside knowledge (in field expertise), or it's just gossip. And of course it can be "easily" estimated how novel the statement is, what quality the sources and support are, and so on.
- jessriedel 9y ago
  The most important change we could make would be to consistently upvote comments that are explicitly abstracts/tldr's, rather than downvoting them based on the mistaken thinking that this will force people to read the entire article and improve discussion. There is latent desire to supply this service for free if it isn't penalized.
  - dfc 9y ago
    Or people could read the articles. For most topics don't want to bother to have a discussion with someone that only knows what was in the TLDR comment. Maybe it's a little too Aristotelian for your tastes but I think taking the time to read the article is a decent barometer for genuine interest in discussion. If I wanted to read lulzy drive by comments on articles based of what was in a TLDR I would not frequent HN.
    - jessriedel 9y ago
      See this subthread for some arguments against that. https://news.ycombinator.com/item?id=15083088 https://news.ycombinator.com/item?id=15083088
      - dfc 9y ago
        Why don't you summarize the arguments for me? Too many words for me on that page.
        jessriedel 9y ago
        Not worth my time, but I'd upvote someone who did ;)
- joncrane 9y ago
  There's one on reddit that does a great job.

Eternal Semptember tho. Give it time.

I've seen an increase in the number of NLP bots that give snippets, but they're far from consistently being accurate. Are there any services that can give you summations like this, kind of like the WSJ's What's News section?

whimsy 9y ago
Coverage isn't great, but these two have managed to get in to my bookmarks: http://thecontext.net/ http://thecontext.net/ https://legiblenews.com/ https://legiblenews.com/ I seem to recall a similar weekly service for HN but I can't recall the name of it.
- j_s 9y ago
  http://n-gate.com/ http://n-gate.com/ - just kidding! Maybe http://www.hackernewsletter.com/ http://www.hackernewsletter.com/ but that's an email.
  - vog 9y ago
    From the n-gate page: > Pasting n-gate content or generally talking about n-gate on "Hacker" "News" is a violation of the Prime Directive. In other words, they want to publish their rants, but are unwilling to discuss their views with anyone, not even the very people they are criticizing. This may be a valid (although insane) approach for a blog, but somewhat contradicts their stated goal: > The purpose of n-gate is to be a remote island of contemptuous sanity in a sea of ridiculous webshit.
  - kronos29296 9y ago
    Is ngate some kind of let's rain shit on hacker news kind of site? Because it looked like that.
    - vog 9y ago
      This description is unfair and not correct. They also rant about FOSDEM.
rangibaby 9y ago
Autotldr on Reddit does a very good job
- lugg 9y ago
  It's not perfect but does the trick 9 out of 10 times, it just uses tfidf irrc.
wtallis 9y ago
Sometimes, the best summary is "this article is bunk". You won't get that from a bot. The fatal flaws in bad journalism are too often subtle and require careful parsing or outside knowledge to identify.
- pas 9y ago
  I think it's easily quantifiable what's a good article. Clear statement at the top that refers to big things, not a storytelling bullshit opening, then the body of the article provides support for that top statement, and later portions are basically just more of the same gradually unfolding sub-articles. If this structure is lacking, it'll be obvious to sentiment analysis, that there are too much free floating statement, and the article either requires too much outside knowledge (in field expertise), or it's just gossip. And of course it can be "easily" estimated how novel the statement is, what quality the sources and support are, and so on.
jessriedel 9y ago
The most important change we could make would be to consistently upvote comments that are explicitly abstracts/tldr's, rather than downvoting them based on the mistaken thinking that this will force people to read the entire article and improve discussion. There is latent desire to supply this service for free if it isn't penalized.
- dfc 9y ago
  Or people could read the articles. For most topics don't want to bother to have a discussion with someone that only knows what was in the TLDR comment. Maybe it's a little too Aristotelian for your tastes but I think taking the time to read the article is a decent barometer for genuine interest in discussion. If I wanted to read lulzy drive by comments on articles based of what was in a TLDR I would not frequent HN.
  - jessriedel 9y ago
    See this subthread for some arguments against that. https://news.ycombinator.com/item?id=15083088 https://news.ycombinator.com/item?id=15083088
    - dfc 9y ago
      Why don't you summarize the arguments for me? Too many words for me on that page.
      - jessriedel 9y ago
        Not worth my time, but I'd upvote someone who did ;)
joncrane 9y ago
There's one on reddit that does a great job.