Hacker News Reader

Comment by lemonspat | Hacker News Reader

You could add a /sarcasm to it for clarity, instead of posting a follow up comment

I think /sarcasm takes something away from it. If I put that there, people don't have to actually think about the comment and what I meant, or their own reaction to it. Using /s tells them automatically, and allows a reflexive response for/against the comment that short-circuits actually thinking about it. I'll risk the downvotes, they don't bother me.

But I am curious about why people downvote, hence my follow up. Clearly some people will downvote a thing they disagree with, or is flame bait. But will they downvote a thing they might agree with, but it also worded as flame bait? That sort of thing. I really am curious about the various behavioral patterns involved in awarding internet points.

/sarcasm is needed on the internet. Not everyone is a native English speaker, and even those who are have very different ideas around humor and written verbal queues.

Good points! I'll keep that in mind in what I hope is already a relatively cautious approach to that sort of comment. On the other hand, ditching all/most idiomatic English and subtleties is probably not the right approach either, or avoiding all cultural references because others might not understand-- I don't think you're suggesting that, it's just my observation that this issue exists on a spectrum and not a binary.

Edit: You're in AI-- what's the landscape look like for improving automated translation to detect sarcasm? As a problem domain, it seems like it would be a subset/cross-section of sentiment analysis & word-sense-disambiguation.

These days the biggest issue is getting a decent labelled dataset.

Once you have that you just fine-tune a BERT-dervived transformer neural network.

See https://paperswithcode.com/task/sarcasm-detection/codeless for some examples and datasets.

Reply on news.ycombinator.com