Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
[flagged] Show HN: Hater News – Find trolls and anyone's hater level on Hacker News (haternews.herokuapp.com)
50 points by KevinMcAlear on Oct 27, 2014 | hide | past | favorite | 53 comments


I got a score of 3.27%.

That said, the methodology here is totally broken. I expected my worst comment to be related to one of the many gender debates on HN, but no.

Apparently, it's this:

----------------------

You can try "Masters of Deception: The Gang that Ruled Cyberspace".

It's a fictionalized retelling of the stories of the two prominent BBS hacker/phreaker groups during the 80s and early 90s.

It's not really brain food or anything like that, but it's a quick and decent read if you have free time.

---------------------

Truly this is the pinnacle of vitriol. I then looked at some of the other darker hues in the periodic table. More mundane stuff.


The last two letters of the "bulletin board system" abbreviation might be what the software is latching on to.


So the Scunthorpe problem is in effect?


Well, a master troll is abrasive using ideas, not words, right?


"The internet is covered in trolls who are filled with hate. It's time we started to expose them and put them in their place."

Can't say I agree with that sentiment. It comes across almost as bad as the hate it proclaims to wish to expose.

I don't think the Internet is covered in trolls, the trolls are an extremely small percentage of users.


This is one of the very few sites I've flagged in my 3+ year HN history. Here's why: HN seems to work because we focus our attention on high-quality posts and "content-filled" comments. Low-quality (and sometimes unpopular, unfortunately) comments are penalized.

This project -- if it does what it claims -- seems to focus on people that create low-quality comments (i.e., "trolls"). In a way, it's rewarding them for their behavior. (and yes, I realize the irony in my comment).

Also, it blanket labels anyone meeting its algorithmic threshold as a "hater", when in fact they may just have written something unpopular.

So I'd prefer for people to create value using HN's API in more positive, constructive ways.

Just my opinion, obviously.


I would say this project alone seems harmless enough, but this regarding this comment from the author's blog post I have to say I agree with you on:

> *Side Note: Eventually I would love to turn this into a Chrome App that will just real-time analyze any user on a page when you visit hacker news and put a score right next to them so the world can see if they hate or love. I’m also going to build versions of this for Twitter, reddit, Instagram, Facebook, and maybe even dating apps. If you want to help, reach out!

http://kevinmcalear.com/thoughts/building-hater-news/


I think both of you have great points. I am not trying to demean anyone or give extra credit to trolls. I do agree that this app is focused on "haters" though. It was more just a fun experiment to see how well it would even work. One thing I noticed is that most people on Hacker News have well thought out comments and are not in fact "haters" (usually only a few percent likely they post insulting comments.) As far as a chrome extension, my intention would be to help give people context for comments based on previous actions. I think transparency can sometimes be powerful and give someone an additional lens to look through when encountering an encouraging or discouraging comment.


Apparently, this is my worst comment... which can't possibly be true:

    "Even then, you're aware of the contacts or the implant, 
    unless you're living in the Matrix and were genetically 
    engineered with a AV jack in your spinal column." 
Only 3.15%, i'm kind of disappointed with myself.


@KevinMcAlear - can you share any details on the algo? :-)


Sure! I wrote a dry blog post about it actually. http://kevinmcalear.com/thoughts/building-hater-news/

@krapp there were some challenges building out a great model but you can download the whole repo and pull out just the machine learning part and see what I did, I have it commented out in an iPython Notebook. :)

The Repo: https://github.com/kevinmcalear/hater_news

It basically uses word tokenization using scikit-learn's count vectorizer and some extra features I added like "bad words", ratio of bad words to total words used, speaking in all CAPS, and a few other features. I then took the features and use logistic regression to predict the likely hood that a specific comment is insulting then average all a user's comments into one score.

I used training data from a kaggle competition and was able to score near the same level as the winners but it will definitely be improved as I keep working on it.


It's open source https://github.com/kevinmcalear/hater_news

Looks like a counted bag of words + https://github.com/kevinmcalear/hater_news/blob/master/its_p... are fed into a logistic regression


Like every other site on the internet with game-ified comment scoring (starting perhaps with Slashdot), it mostly encourages groupthink and discourages dissenting views from being expressed.

OTOH, at least HN lets you create throwaway accounts easily.


1.7% for me. It seems to show constructive criticism as "hate," though I bet that's very hard to get perfect.

I like the little time waster videos and whatnot that show when it's taking a little while to compute.


Score: 4.65%

Worst Comment: "You're just making a caricature of yourself."


  An error occurred in the application and your page could not be served. Please try again in a few moments.
aww I always get here too late for the cool stuff.


Should be fixed now! :)


It is! My most hateful comment is about the Apple Air: "Please provide an example why a screw or removable panel would provide less value to the consumer than gluing the thing and making it impossible to replace the battery on your own. I eagerly await your response."


Why, your seventh word is obscene language. I dare not repeat it because I don't want to be branded a hater by someone's glorified grep program. :)


Reliability. Screws come loose, especially when flexed in a pocket etc. Its considered a win in mechanical engineering circles when you remove a screw from a design.

I look forward to metal-air batteries, where the charge can be expected to exceed the lifetime of the device.


Not to run too far away from the actual content of the post but...

I thought metal-air batteries didn't produce a high enough current to be used in devices requiring much more juice than a hearing aid?


Alcoa runs a car on an aluminum=>alumina metal-air battery!


Even when the part is normally a replaceable part? Perhaps this explains the increase in difficulty in repairing cars nowadays.


Lose the sound, please- or at least warn people so they can turn off their speakers in an office.


It is not.


Just got that error.


It couldn't find me on HN then ran some auto play video.


Sorry! The site has been getting hammered with people so it keeps crashing. Trying to scale it up now. It has to make a separate API calls for each of your comments (max 50) so it takes a while and since there are so many people using it sometimes it times out...


Background jobs are good for this sort of thing. Hope it's not being done in a controller...


Interesting, my most downvoted comment is also my highest rated: "Downvoted because enough is enough! The "14 competing standards" comic was really funny the first few hundred times it was snarked into HN comments." https://news.ycombinator.com/item?id=8354087


The worst comment: "How do you know Uber drivers use iPhones exclusively?" -tinkerrr Can't believe I spewed such hate


Same here:

"If you have multiple modes you can basically reuse the entire keyboard for commands. That's one advantage that is impossible to achieve in non-modal editors. Ctrl-ing, Alt-ing, Shift-ing stuff just isn't the same, especially as those commands get more and more complex."


From the blog entry[1]

  Some Final Thoughts.
  - Sarcasm is hard to detect.
Reminded me of park or bird[2].

[1] http://kevinmcalear.com/thoughts/building-hater-news/

[2] http://xkcd.com/1425/


haha. :)


pg got a 2.95% with worst comment:

"The sort of incivility we're worried about is the more explicit type where someone replies to a comment with "You are an idiot. Don't you realize that x y z?" when they could have simply said "x y z." I'm pretty ..."

It would be cool to ignore/weight down quoted text somehow.


I bet you that the algorithm could be adjusted to consider comments that meet the following criteria:

(1) If user has the most upvoted comment in any thread (2) Check if that thread had a lot of flamewar language.

If so, return that as the most hateful comment. I'm only half joking about using this approach.


On the other thread I said that it seems to be counting quoted text as my words.

It's a fun little app! Not sure how useful it is, but I do want to see how it develops.

I'd be really interested to see if there's any link between HN Avg karma and hn hater score.


@KevinMcAlear. I found a serious flaw in your algorithm's results.

Your own hate score is 3.61% Should not this post and the creation of Hater News itself up your score to 100%?

My own score is an even lower 2.06%. What gives? Maybe this comment will give me some points.


Clicking on the text of the Worst Comment actually takes you to a different comment. :-)


I'm getting an error message. Stupid app!

...So did my rating just go up?


It would be nice to have the username used in a search added to the url of the results page, to make the results easily shareable.

Might have a look-see at the source.


'The Worst Comment "Been looking for something like this! Perfect. Thanks for the effort!"' - right. sorry about that one.


> Count Vecteriztor

took me a bit to understand that was a weird typo and not a really cool CS concept.


Also s/Logisitc/Logistic


Whoops! Fixed! Thanks guys.


> Application Error

I hate hater score sites that don't give me my hater sore...


Getting a mix of application errors and 400 errors.


commented the first time this made FP, not sure what happened. I still think this is a cool use of the HN API.


This already made the front page once today before it was nearly at the same moment

1) Taken down via HN hug of death

2) Nuked from orbit by mods.

Also I'm only at 2.5% I honestly figured I was far more toxic.


> Nuked from orbit by mods.

No moderator touched that post. It was killed by user flags. (Also, why on earth would we "nuke" a post sharing original work done with the HN API? That's why we made the API in the first place.)

https://news.ycombinator.com/item?id=8515160

Reposts of user-flagged items, though, we do tend to penalize. Otherwise flags wouldn't mean very much.


I assume flag killed was something mods could do :x I know they can super flag posts, I assumed that is what [Flag Killed] meant.


[flagkilled] means that a post was killed by user flags. I got tired of typing "This post was killed by user flags."


haternews.herokuapp.com application is down. lol should have used javascript.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: