Hacker News Reader

Comment by dcposch | Hacker News Reader

Just an idea: sentiment analysis has gotten pretty accurate.

What if HN included sentiment in the sort algorithm?

The post score you see would stay the same (upvotes minus downvotes). You'd just multiply the sort score by, for example, 1.25 for posts positive in tone and 0.8 for posts with negative tone.

So if two comments have about the same number of upvotes, the nicer one appears higher up.

Criticism or disapproval could still become top post with enough upvotes, but it wouldnt happen as often as it does now.

View on HN

That presumes that the negative comments are always the ones derailing the discussion. For example:

"I think these questions at IAC are great, it's good that people get a chance to <self promote thing> at a technical conference"

"Wtf is with these stupid questions at IAC? They should be asking about <fascinating technical detail>"

Hah, you just described the sheer drivel which is comments on LinkedIn stories. Good point - contentious and _well presented_ negative arguments are still helpful.

Then perhaps use sentiment analysis to remove the derailing part of both negative and negative comments?

i.e.,

"I think these questions at IAC are great,..."

"...they should be asking about <fascinating technical detail>"

> What if HN just automatically penalized negative posts?

"I think it's a good idea to start a national wingsuit base jumping team."

"I don't know if that's a good idea. Pretty high chance of death"

Penalized - negative post

I am currently doing a (simple) sentiment analysis for the complete reddit comment set. Perhaps when I am done with that, I can have a look at the HN corpus (where to download?) for correlations between upvotes and users, though I doubt that rating post quality by sentiment is easy.

But it could be useful to look at the comment history of a user. When someone who is an overall positively minded contributor (with a high average positive sentiment) suddenly writes one scathing reply, this could mark a potentially interesting thread.

You can query the HN corpus with Google's Bigquery: https://cloud.google.com/bigquery/public-data/hacker-news

If that worked at all, it could be applied before the comment was even posted. Warn the user that their comment looks inordinately negative, and ask if they would like to revise before submitting.

This is not about censoring or discouraging criticism. Often criticism is fair and warranted.

I just want positive comments to appear as the top comment more often.

Negative commenta should have a bit of a higher bar (in other words, need more votes) to get to the top

Reply on news.ycombinator.com