[Air-L] Air-L Digest, Vol 175, Issue 17

Sugar, Benjamin N bsugar at gatech.edu
Thu Feb 21 16:00:18 PST 2019


Hi Yosem and Sweta,


1. What is the best method for collecting pre- and post-disaster data on
social media?
2. Is it possible to collect data not just on Twitter but also on all
social media platforms (e.g., Facebook, Instagram, LinkedIn)?

These are sort of connected.  I don’t know anything about Facebook.  I believe that Instagram no longer has an API (or at least not one that people I know find adequate to research with), not sure about LinkedIn.

Another place that I think would be good is Wikipedia.  Here are two examples of researchers who used it to understand events of the type you are talking about.

Twyman, Marlon, Brian C. Keegan, and Aaron Shaw. 2017. Black Lives Matter in Wikipedia: Collaboration and collective memory around online social movements<https://doi.org/10.1145/2998181.2998232>. In Proceedings of the 20th ACM Conference on Computer-Supported Cooperative Work & Social Computing (CSCW '17).

[J2] Keegan, B., Gergle, D., Contractor, N. (2013) “Hot off the Wiki: Structure and Dynamics of Wikipedia’s Coverage of Breaking News Events.”<http://brianckeegan.com/papers/ABS_2013.pdf> American Behavioral Scientist, 57(5): 595-622.

Finally, there’s Reddit.  I think Google’s “big query” maintains the continually updated archive for Reddit.

As for Twitter, the problem is that you can only go a week back in the timeline. There might be a dataset available.  Two good places to check are the Harvard Dataverse and the Internet Archive (for which I think the Dataverse is includes).

The other thing you can do is search for papers on the topic and then write and ask if they’ll share their data.  Two (of I’m sure many) that I know of are:

http://faculty.washington.edu/kstarbi/

https://cmci.colorado.edu/~palen/papers.html

3. Which social network analysis (SNA) tool has the easiest learning
curve and is the easiest to use to analyze relational structures on social
media?

Assuming by relational structures you mean network statistics, then definitely Gephi.  It’s sort of the “photoshop” of network analysis.  Just like Photoshop, it won’t be readily apparent how to do even simple things at first, but there are lots of tutorials.

If you feel comfortable with basic programming and reading documentation, then for simple things I think NetworkX is a good choice (there’s also a javascript version of it as well I think).   Cytoscape (and cytoscape.js) is good, too but I don’t find it as easy to use.

Happy to answer any other questions!

Benjamin


More information about the Air-L mailing list