Heatmap of Twitter knowledge by location (heatmap.js + Google Maps) | Hacker News Reader

Heatmap of Twitter knowledge by location (heatmap.js + Google Maps)

inboxq.com

35 pointsdrpancake25 comments

Neat, but on just about every topic there's a huge bias towards California, North East Coast, and England (because those are tech hubs?).

There also seems to be a contingent of knowledgeable Twitter users in a boat off the west coast of Africa. (I'm kidding - this is due to tweets where lat,lon = (0,0))

The Nigerian scammers are migrating to Twitter, and have taken up seasteading to avoid the authorities.

Agreed. This would be way cooler if it took total tweet density into consideration so that it wasn't always London, NY and LA.

Definitely something we've considered for a future enhancement. Our summer intern worked on this map over the last few days and we wanted to release it before he headed back to the UK later this week.

I wouldn't mind getting my hands on a summer intern who could turn that out in a few days.

James is awesome. And, we found him via Hacker News!

Try sports teams. "Broncos" creates a huge hot spot over Denver. They are a good way to validate the data/app generally works.

Tried a few German sports teams. "Schalke" was a huge hot spot over London... huh.

We only include accounts that Tweet primarily (really exclusively for practical purposes) in English. So it may be that there are more English language accounts in the UK that are fans of that team than in Germany. Just a guess. I think the various data normalization and other enhancements discussed throughout this thread would also help.

except the Jesus topic, because of all the godless heathens in SF and NYC

Population size needs to be factored in. NYC, LA/SF, Boston, London, Chicago, and a couple other large cities are pretty much always the top, primarily because they are the largest English speaking population centers.

Might be nice if the map didn't hide part of the world (Australia, New Zealand etc)

Sad to see that Australia didn't make the cut :(

It looks like they're fetching a static image from Google Maps and using a jQuery library to do the heatmap (which uses canvas). Not sure why they made it that way. Would have been a perfect fit if they fetched a rectangular map and got moved that text on the right.

In fact the portion of the map shown includes a minority of the world's population.

It would be great to do something like this for an election. A heatmap for people who voted #red or #blue. Seems like a lot of neat uses for this.

On topics where there's not many tweets (e.g. haskell), it's nearly impossible to tell that anything on the map changed at all. Perhaps the data also needs to be normalized?

Yeah it needs a few things done before it can become usable. Get rid of the assumption knowledge = frequency of keyword. Normalize the data. Account for population differences.

Would also be nice if you could pan around but it looks like the code relies on a static image.

Thanks for the great suggestions. All are features that we'll add in the future. The goal was really to get something out there that had a small bit of utility and spend more time on it if people found the visualization interesting. As I mentioned a few threads up, this is something that our outgoing summer intern worked on during his last few days with us and we wanted to release it before he headed back to school.

You're most welcome! I hope I didn't come off too harsh - I understand it's an early preview as most things are. I look forward to seeing it flourish.

Nice idea, but no queries I could think of yielded any surprises, and almost every topic was dominated by the media capitals of NY, LA, and London.

Apparently, also, there is no expertise worth plotting anywhere in Asia or even Australia.

It was more fun to type in diseases and pretend the overlay was a map of contagion. EG: [swine flu]... There's a massive outbreak 1500 miles west of the Galapagos!

I hunted around for a few things like "node.js", "Php", "True Blood", "Harry Potter", etc... they all fit within the same pattern you noted. (with minor, or not-unexpected variances)

My favorite was "Maple Syrup"

Gotta wonder what the defn of 'knowledge' is here - just mentions of the keyword? Does that indicate knowledge?

No- it's not just keyword mentions. Our main product analyzes the both the content of an individuals Tweets and social factors to make decisions about what they know. I'd be happy to share more about how we do that- shoot me a note at joe@inboxq.com if you want to chat.

You need to normalize the results by population density.

Pretty much anything I type, I get the same results - NY + LA + SF + London.

Reply on news.ycombinator.com