Live data from Hacker News

The one million tweet map

onemilliontweetmap.com

21–30 of 36 posts

Re: The one million tweet map

#22
post #18

There is quite an intense location on west coast of Africa, a bit below Nigeria that appears to be coming from somewhere in the ocean, just to the west of Sao Tome. Google maps is not showing me anything from that location like an island. Any idea why this might be? Also useful would be some sort of Distance scale/measure.

Looks like 0 degree latitude, 0 degree longitude. Probably tweets with no location info or bad location info.

Re: The one million tweet map

#27
post #22
post #18

There is quite an intense location on west coast of Africa, a bit below Nigeria that appears to be coming from somewhere in the ocean, just to the west of Sao Tome. Google maps is not showing me anything from that location like an island. Any idea why this might be? Also useful would be some sort of Distance scale/measure.

Looks like 0 degree latitude, 0 degree longitude. Probably tweets with no location info or bad location info.

That is Null Island

http://www.nullisland.com/

Re: The one million tweet map

#28

Earlier quoted context omitted.

What do you mean? We just display data from twitter stream. You can use their stream if you want to

Which stream are you using exactly. Are you applying any filters? Afaik the complete firehose has restricted access, or did that change? Edit: very cool visualization btw. What are you using exactly on the backend for stream processing?

I use the public stream with a filter on location (:locations => '-180,-90,180,90') Twitter doesn't seem to have limitation for that.

The backend is written in Ruby. The daemon to index tweets use the great Intridea gem https://github.com/intridea/tweetstream.

The page http://onemilliontweetmap.com/ is simple Sinatra application, and I use REDIS pub/sub to push tweets to the browser.

Re: The one million tweet map

#29
Very nice and more interesting than the bar charts/graph that other twitter visualization tools give us. It would be really cool (and informative) if you could add a small word cloud when the cursor hovers over a city.

I did a similar non-interactive dataviz a while ago during the big student protests in Quebec this spring. I clustered (using LDA) the tweets talking about the protests and I mapped and identified them.

http://olihb.com/?p=199 (in French, though)

Post reply on HN