Ask HN: What do I do next? | Hacker News Reader

Ask HN: What do I do next?

2 pointsmysteryleo11 comments

For kicks and giggles (and honestly to keep my mind sharp), I built a memetracker. It is a good mental vacation from the day job.

It leveraged a lot of my technical interests in algorithms and machine learning. I put the result on the web, and am watching my statcounter flat line. Well, I get some hiccups when my gf visits.

So I guess the internet isn't a "build it and they will come" kind of place. Reality Checks are hard.

I'm not sure what I should be doing next to get the word out. Is there some magic I should be doing. Any veterans out there know what I should do?

Advice and pointers would help.

View on HN

The site itself doesn't have content, it has a content aggregation service. That makes it hard to spread in a viral fashion, especially when it's an untargeted service spread amongst many verticals. It's useful, but the "wow check this out" factor is missing.

One improvement would be to add a comment system and rely on UGC to keep people around. But then you have to devise a way to populate stories and make the site look livelier than it is, which may require some kind of comment bot.

Another thing you could do is start drilling down to smaller and smaller niches and try to become a top search result in those, and then work your way back to the huge verticals you currently have. As it is, you are basically competing head-to-head with Google News.

Yeah, we don't have content ourselves, our value add is the ranking and filter service. Also, allowing you to see the conversation across the web.

UGC would be tough, because I'm at a chicken and egg state right now as you pointed out. And crap comments would lead to more crap comments. Imagine if Quora or Hacker News had a comment bot. Sets a bad tone for the future.

i guess there may be a flaw in my initial strategy.

when training my algorithm (and as i make iterations), I compare the results to google news to make sure we're as good if not better. That's why a lot of the initial top-level categories match up pretty closely.

Sports vs sports http://sports.rawsignal.com/ http://news.google.com/news/section?pz=1&cf=all&ned=...

However, once I'm more confident in the algo, I'll let it loose on something more niche, like college football

http://collegefb.rawsignal.com/

that's how we can differentiate from google, but i never realized that having too many options would paralyze people.

maybe if i release each vertical under its own domain . . .

I think that nailing down the smaller more focused categories is the way to go. If you can make it both more relevant and more focused it would be both different and useful providing a real reason to give it a try.

sure i can work on that next. what initial categories would rock your world?

also, should i brand each category as its own site, or is the current model ok?

ok, i put up baseball, basketball, and football:

http://football.rawsignal.com/ http://basketball.rawsignal.com/ http://collegefb.rawsignal.com/ http://baseball.rawsignal.com/

had to refungle some code to speed up our analysis, but it should be working. hope you like it.

First step: share a link.

what are good places to share a link?

The main vertical I'm focused on right now is tech: http://tech.rawsignal.com/

although I built other verticals that seem to be working already. Not sure where to share it that would interest people who like tech.

Right here.

Regards

I'd love to be able to filter or have the system filter even further. For example in sports, just baseball or football not all sports.

I do see that you have something that sounds like this coming under your coming soon section so you may already be working on this.

yeah. the algorithm I built takes a little training per topic/vertical, but after the initial training, it just starts learning.

I have a long list of potential topics that the algorithm would be good at. pretty much any sports topic it would kick butt on.

My brother keeps telling me I need to split up sports too. Basically, the initial verticals were to stress out the algorithm to make sure it would work on various concepts and still keep things interesting.

Basically, you could imagine every concept http://alltop.com/ tracks running through my algorithms.

Thanks for the feedback though. That's the kind of information that really helps a hacker getting started.

Reply on news.ycombinator.com