News for hackers. Stuff that matters.
Re:A curated list of FOSS tools to improve the Hacker News e... by minimaxir on Monday March 15, 2021 @12:21PM (#26467937)
In natural language processing and Hacker News in particular, it's often incorrect that a submission with mostly positive comments imply an article is valuable, and conversely a submission with mostly negative comments imply an article is not valuable.

That rationale just leads to even more groupthink than usual for an online community.

Re:A curated list of FOSS tools to improve the Hacker News e... by creamynebula on Monday March 15, 2021 @12:30PM (#26468080)
Maybe the quality of submission is not what's being judged, but whether or not the discussion itself is valuable. I for one prefer to read discussions in a positive tone...
Re:A curated list of FOSS tools to improve the Hacker News e... by clairity on Monday March 15, 2021 @01:44PM (#26469049)
more pointedly, comment sentiment is a naive and misleading signal of either submission or discussion quality. most simply because it's a poor proxy for why, but also because quality is an amorphous concept that defies direct rationalization (i.e., being numericalized), among other factors.

naive heuristics are also gamed more easily, especially because they have a tenous relationship to the desired signal in the first place.

slightly more interesting would be absolute value of comment sentiment, which would be a (still naive) measure of controversy/engagement. to really get at quality and value, you'd have to consider comment semiotics/semantics (the symbolic meaning and content), which complicates the effort exponentially (perhaps impossibly so).

Re:A curated list of FOSS tools to improve the Hacker News e... by invisiblerobot on Monday March 15, 2021 @06:31PM (#26470152)
That's not what I'm saying. I'm saying an article that is positive sentiment OR negative sentiment is INTERESTING enough to click on. An article with luke warm sentiment either way is less interesting.