11–20 of 43 posts
Er, my "most insulting" comment is this: https://news.ycombinator.com/item?id=8293794 I guess the algos need more training, which I'm helpfully providing with this comment.
https://github.com/kevinmcalear/hater_news
I like the idea of scoring people on their "troll" number. This might fix other conversation sites where there's lots of noise. Someone could sell the solution to the Wall Street Journal.
"Ah ok. I looked at it more as something you force on your co-workers :)"
Looks like 'ah' implies negativity to the algorithms driving this.
Edit: It's also only pulling your 50 most recent comments by default. That covers about the last two weeks for me and is probably why it isn't very accurate. 50 comments is a very small dataset to work from. The other option it gives is 50 oldest comments - again not very representative for someone who has been commenting regularly for 1301 days. I try to keep my comments useful but I know for a fact I've gotten into emotional debates in the last 4 years in which I was negative/stupid/hating. Maybe it's not possible but I would suggest pulling a random sample of 100-200 comments for more accurate data.
Not to hate, but this is really not working
Actually disappointed I scored so low on this. I think it's healthy for community discussion that HN users to have sizable hater scores - otherwise we're a bunch of yes people (like me, it seems).
This is stupid.
I would be more interested in measuring how often a person up/down votes, how often they are up/down voted compared to others within a similar posting. It is one thing to get down voted into oblivion but if you have a lot of company does that mitigate it?
"Ah, fond memories of downloading MP3's via irc DCC :)" worst comment?