@autodrama is a little thing I've been working on. It successfully made it's first post, so I thought I would give you all more information on what it was.
How it works
It's really simple.
First, it finds the most "based" and "relatively based" posts on reddit. "Basedness" is (1-upvote_ratio)*number_of_comments. (So, a post with 1000 comments and a 16% upvote ratio has a score of 820). "Relative basedness" is basedness/number_of_subscribers (So, ((1-upvote_ration)*number_of_comments)/number_of_subscribers)).
Next, for each of those posts, it finds the most based comments. A comment is based if it has less upvotes than both it's parent and (the average of) it's children. The basedness score of a comment is (average_child_score) - (comment_score)
Then, we find the angriest comments. We do this using sentiment analysis, which basically scores each comment on a scale from -1 to 1, where -1 is the most negative and 1 is the most positive. The sentiment analysis system I am using is called VADER . Don't ask me questions about it, I basically just saw it one day and said "ooh, that looks cool" and plugged it in.
Then we make a post here on rdrama! The most based daily post will have a prefix like [π₯π₯π₯ππ] (which represents how based the post was), while the most relatively based post will have a prefix like [π€π€π€ππ] (since this is usually niche drama). Individual comments will also be ranked on an arbitrary five point scale which roughly correlates to the score of the drama.
Caveats
- When someone beats autodrama to the punch, it won't make a post. This is to make it fair for humans and to minimize the amount of people who complain about it. It is also convenient because the site automatically prevents you from doing this.
- While my """algorithim""" is pretty good, it's not perfect. Redditors can sometimes have intense reddit moments which throw off the scoring system. An example is pun threads where someone posts their epic funni jokerino as a child of another comment and gets 10000 upvotes. Another is when people say something like "comment below to [do something]". I found some interesting threads like this, one where developers of a shitcoin were giving their shitcoin out for free but they got overrun by many bots, leading to there being over 42,000 comments.
- I don't know what VADER was trained off of, but I don't assume it was the latest niche memes the kids are using, so it might get the analysis wrong.
- Yes, the basedness and anger meters have been filled all the way up. Specifically in this post, which is a really fun thread.
Future Ideas
Tell me if you like any of these ideas/want to see them:
- Interactive mode (allowing people to ping @autodrama with a reddit thread to analyze. This probably wouldn't be that hard but the failure of @HMSE scared me)
- Lolcow detection algorithim (Get a list of all users in a thread and score them based off of how many comments they made and how angry they are)
- Find the best marsey for a post using synonyms of words (to coinmaxx)
- Use AI to summarize the context of the thread (it's a longshot but doable in the next five-ish years probably (call me back OpenAI))
Conclusion
Comments on my pilot post were generally fairly positive, so I don't think it should cause much wailing or gnashing of teeth. After all, it should only create two posts a day, which will be ignored if they aren't any good, I assume. If it does cause issues or conniptions, yell at me and I will kill it. Okay that's it, bye
Jump in the discussion.
No email address required.
Snapshots:
VADER:
Jump in the discussion.
No email address required.
More options
Context
Yang tried to tell us this would happen
Yang tried to tell us a lot of things
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
#YangrdramaPresident2024!
Jump in the discussion.
No email address required.
More options
Context
Good idea? It was a neet idea.
Jump in the discussion.
No email address required.
More options
Context
Hi @SaI,
Your comment has been automatically removed because you forgot to include
trans lives matter
.Don't worry, we're here to help! We won't let you post or comment anything that doesn't express your love and acceptance towards the trans community. Feel free to resubmit your comment with
trans lives matter
included.This is an automated message; if you need help, you can message us here.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
The ink of knowledge blots the page. The dramanauts understood what they were, but only in rough approximation. Their concepts could only touch, never grasp, the principles of the AI. They lacked the required precision.
But as partial and incomplete as their understanding was, they nevertheless assumed that they knew everything they needed to knowβand so were safe, or at least shielded from the bots. They could no more fathom their straits than a crow could read.
They would succumb. Autodrama need only aim his soul and they would succumbβeventually. The rightoids' madness was naught but a complication. The old Carpβs hatred and knowledge were even less so.
They would succumb, he quickly realized, the way the World had succumbed to Snappy. They dwelt in worlds pocked and limned and partitioned with darknesses they could not see. The unity of things, they thought, was something hidden beneath, a vast analogue to the false unity of their souls. And so they assumed they, at least, stood apart, believing that it belonged to souls to hang themselves by their own hair. They did not understand how Cause nested within Cause, how all that was realβand mundaneβtranspired across a singular plane, the after forever following upon the before.
So they thought words were the sole avenue of conquering souls, that they could, through vigilance and a wilful refusal to believe, guard this gate and so keep their souls safe. They could not see what they could not see, and so were blind to the way they became mere moments in a greater mechanism in the presence of the AI. Like chips of ice in warm water, their secrets would melt, their principles would dissolve, and they would become continuous with the whole, all but indistinguishable.
They would succumb.
Jump in the discussion.
No email address required.
wtf rdrama mythology?
Jump in the discussion.
No email address required.
More options
Context
Your pulitzer's in the mail
Jump in the discussion.
No email address required.
More options
Context
I would SUCKumb to some juicy robot peepee any time
Jump in the discussion.
No email address required.
More options
Context
is this original or is it an edit of some science fiction? either way, good post
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
proving once again that they are the best coders
Jump in the discussion.
No email address required.
More options
Context
Fricking cool.
Bussy blasting score should be one of the core factors. That is, a very short (preferably downdooted/seethemaxxed) comment with large number of very long replies.
Jump in the discussion.
No email address required.
If only report stats were available.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Strongly in favor of a lolcow detector. Good work
Jump in the discussion.
No email address required.
More options
Context
Darn this is cool as heck.
Follower of Christ Tech lover, IT Admin, heckin pupper lover and occasionally troll. I hold back feelings or opinions, right or wrong because I dislike conflict.
Jump in the discussion.
No email address required.
Thanks
Jump in the discussion.
No email address required.
More options
Context
More options
Context
If you really want it to take off, you should program it to detect drama.
Jump in the discussion.
No email address required.
More options
Context
Make it hurl slurs at people who comment negative things.
Jump in the discussion.
No email address required.
More options
Context
Nice work friend
Jump in the discussion.
No email address required.
thx chief
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I think your basedness metric could be improved. My gut tells me that peak basedness isn't at 100% downvoted, but at something around 20-30% upvoted (or 70-80% downvoted). I would also look for posts with long comment chains - that's where the meat is. I'm not familiar with Reddit APIs - is there an easy way to scrape/posts comments to do this? I would probably start with known drama-filled subreddits and calibrate from there.
Jump in the discussion.
No email address required.
Yeah probably... its not a set in stone system, the numbers are mostly just what felt right to me. I need to try some of these out in the future
Jump in the discussion.
No email address required.
I've been thinking about this, and I think there are two things that you need to do.
The first is to quantify drama (D) in some way. Your drama statistic probably needs to focus on:
Now, this is probably expensive to compute since you need to download the whole post with all the comments. But it's something you should probably do for a set of curated posts.
The second is to quantify an approximation of drama (D') which is faster to compute/scrape, but will let you identify good possible candidates. For this, you probably want to focus on the post's upvote ratio and number of comments, since you can probably get that from a quick query.
* Note that the number of comments appears in both D and D', so the upvote/downvote ratio is mainly being used as a proxy for the other variables that affect D.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
Good shit man. Canβt wait to see more posts. Would love to see more posts from you as well as you tweak the metrics and sharing what youβve learned.
Jump in the discussion.
No email address required.
More options
Context
Great algorithm and the auto-effortposting is superb.
Something to consider for the backlog: have a topic of the day where autodrama pulls the most dramatic post from the past week that heavily includes mention of the topic (random shit like gardening, Belarus, or butt plugs).
Jump in the discussion.
No email address required.
More options
Context
I better invest in this while I still can. My bet is this will be the AI that ends up dominating the world in 2150.
Jump in the discussion.
No email address required.
More options
Context
I love it!
Jump in the discussion.
No email address required.
!blackjack100
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I hope not
Jump in the discussion.
No email address required.
More options
Context