Hacker News Reader

Comment by jrumbut | Hacker News Reader

A very cool demo and I congratulate the author, but I am always a little sad for more data science type demos that try to answer the question (that is proving toxic) "given what I know about you, how can I find a community of people just like you?"

I would love to see a subreddit finder that answers questions like "what community would complement your interests?" or "what community needs to hear what you have to say?" or "what community would be made better by your presence?". Similarity is at best a proxy for it.

Those are harder but, I think, more useful.

View on HN

Thanks for your valuable feedback. This indeed answers a question such us "What community is used to hear what you have to say?".

It is not really based on your interests, it just takes your text and suggests subreddits where people have posted similar texts.

Otherwise, I agree on what you say. I would love to also see those kind of systems. Kind of what you get as a reaction when you talk with a mentor that surprises you ;)

Unfortunately, my skills are not there yet but I am working hard to eventually be able to build those "surprise/discovery" systems.

They aren't plug 'n play for advertising, though. "How can I find a community for you," is the charitable flipside to, "here's a community you might like to be a part of," where the community is "Coors Lite purchasers."

That's so short sighted though; I'd be so much more likely to engage with an advert that figured out a new thing that would interest me than the usual "you've been reading about sc2 for a week, so have more of the same" nonsense.

Yeah nobody knows how to do that. To the degree that anybody's figured any amount of it out, it's much more likely and lucrative to point that code at changing your vote than your brand of toilet paper (wink wink).

Good point, but I think the goal that you outline is harder and that this type of research is a stepping stone along that path.

Reply on news.ycombinator.com