Time. Location. Emotion.
by: Paul Saarinen
It’s about 4 or 5 PM last Thursday, and I’m sitting on my couch at home suffering from a low grade fever of 101 degrees, and a thought occurs to me. Well, it’s the thought illustrated in my little diagram above. I always tend to think of data, and how it can be used through a marketing filter. This is my first of many problems. In order to try and understand the diagram above, I had to forget those thoughts…at least for a few hours.
We now can bring three seemingly unconnected, but readily available data points together, and look for connections in a whole new layer. Are there implications, or will anything be on the other side? I’m trying to get my small brain to wrap around something much larger than I can swallow. I guess what I’m trying to peer into, is a collective consciousness, and how it relates to real world events.
Step 1. Gather Twitter api search feed.
Step 2. Utilize GPS coordinates for tweets that have them, or sub in location profile data for those that don’t have GPS data.
Step 3. Parse each 140 character tweet for emotional trigger word and give a value (this will be tricky, and pretty loose)
Step 4. Use open source heat map API synced with Google Maps API to visualize impact
Step 5. Start looking for patterns with real world events (I have a few ideas on this)
Step 6. Determine if the patterns prove out over multiple occurences
Well, that’s as far as I have gone. What are your thoughts?
8 Comments, Comment or Ping
FalkinInvesting
That is quite a concept! It would be really cool to see the impact of breaking news and how it spiders out!
Mar 9th, 2009
taulpaul
Thanks for the thought. I was thinking of that a little bit too. The first thought in my head was if there was any tie between murder rates in certain locations, and a measured sentiment in that area before the murder occurs. Maybe it wasn’t my first thought, but I know that data is readily available in a type GIS format, so it’s easier to analyze.
The question I want to pose, can collective emotion predict actual events? Is the twitter base a large enough sample size?
Mar 9th, 2009
Joseph Rueter
Have you seen? http://wefeelfine.org/
Mar 9th, 2009
taulpaul
When wefeelfine.org first came out, I found it fascinating. It does a beautiful job at looking at this type of data in a visceral way. One of the things it doesn’t do well, is quantify with flexibility. Qualitative vs. Quantitative….though Tufte would be proud.
Mar 9th, 2009
cam
I have a connection (somewhere) to a market research group that created an entire emotion implication tool based on word choice. I gotta find that guy.
I smell a mashup.
Mar 9th, 2009
Mark Wagner
Joseph,
The Johnathan Harris thing is what got Paul and I started on the discussion. Paul was the first to point it out (wefeelfine.org)and a few months later I caught up to Paul’s original stream of thought. It is definitely where things are evolving.
Mar 9th, 2009
Adrian Ho
Check these guys out. We have been talking to a bunch of clients about this technology, it has a bunch of applications
http://biomapping.net/
Mar 10th, 2009
Tyler Hayes
Interesting that Adrian brought up the Biomapping resource, I’d never seen that. Inetium did some heat mapping at the National Civic Summit in Minneapolis this summer, though it was not emotion-centric, just simply based on tweets and where they were coming from. Biomapping is interesting, but I don’t think I’d be down to wear a Galvanic Skin Response detector just to add my own data. Unless maybe it was integrated into the back of a watch?
WeFeelFine is so cool, his TED Talk is even better. But I agree with Paul, it needs more description and differentiation within itself. The first time I visited it (stumbled on it, hadn’t seen the TED Talk yet) I was unbelievably confused as to what I was looking at.
As for Paul’s idea, the intent is phenomenal. Not too far from what Google does by mapping how much social mention there is of things like “swine flu” so that people can track outbreaks. My first suggestion: just built it already. And then tweak it continuously. I have yet to see anyone else do this. My only other suggestion: try to find other ways of finding people’s location-based data, like (in the background) searching to see if they also have a Brightkite profile, and looking for their most recent posted location. Or, see if Twitter’s API lets you see IP addresses (or at least regions) rather than their profile location which is sometimes misleading, outdated, or purposely incorrect (i.e. some people put Twin Cities instead of Minneapolis).
Aug 31st, 2009
Reply to “Time. Location. Emotion.”