In my experience having run a bunch of different sentiment models on HN comments, HN comments tend to place around neutral to slightly negative as a whole, even when I perceive the thread to be okay. However I've noticed a huge bump in negative sentiment on large HN threads. I generally find that absolute sentiment doesn't work in most corpuses because the model reflects its training set's sentiment labels. I generally find relative sentiment to be a lot more useful. I have yet to do a temporal sentiment analysis on HN but I have a suspicion that it's gotten more negative over time. I agree with another poster that I think HN needs to be careful to not become so negative that it just becomes an anger echo.
Relative sentiment on this site between topics is something I've done and the obvious results show. Crypto threads are by-and-large negative, most political and news related threads are also highly negative.
https://journals.sagepub.com/doi/pdf/10.1177/014616721878319...
Sure, there's the 20% of comments that are outright rude, or tie everything back to their pet grievance (job satisfaction, government surveillance, the existence of JS).
But beyond that, the technical conversation has a negative, critical edge. A lot of comments come from the angle "You did something wrong by...", or only reply to correct.
There are still golden comments, and most personal anecdotes are treated respectfully, but it makes for an intimidating environment.
Good job doing a whole psychoanalysis based on what's basically a joke, though.
> Good job doing a whole psychoanalysis based on what's basically a joke, though.
Guess there's still some work to be done on that positive sentiment replying eh? :)