Limitations

SeeTweet is by no means perfect, so I recommend that you keep the following caveats in mind when using it to avoid drawing stronger conclusions than the data merits.

The location data is noisy.
Locations are user-specified information, which may be imprecise, inaccurate, or fictional ("Hogwarts" is the location of a lot of Twitterers). They may be out-of-date, because people do not necessarily change them as they travel or move. For instance, I list my own as San Diego even though I am in San Jose half the time. Furthermore, Google Maps (or the SeeTweet code itself) will sometimes misinterpret locations. Google Maps finds "Mars", for example, to be about 20 miles outside of Albuquerque.

Twitter users are not uniformly distributed.
Much as the population of the U.S. is not uniformly distributed, neither are Twitter users; they tend to clump up in cities, especially young cities. A map showing usage of a certain word or phrase in New England but not North Dakota may be due to the relative lack of Twitter users in North Dakota rather than a lack of users of that word or phrase there. You can get an approximate baseline by searching for a neutral word like "the", or by looking at a map of general Twitter usage, like this one by Eric Fischer.

There isn't any negative data.
SeeTweet provides only positive data. If tweets do not show up in an area, this doesn't necessarily mean that people don't use the word or phrase, only that they're not using it on Twitter right now. Negative evidence can be simulated by searching for an alternative to see if that is used in the area. For an example of this, see pop vs. soda on the SeeTweet Tumblr.


Return to SeeTweet