Quantifying Salient Concepts Discussed in Social Media Content: An Analysis of Tweets Posted by ISIS Fangirls
Shadi Ghajar-Khosravi, Peter J. Kwantes, Natalia Derbentseva, Laura Huey · 2016
In this paper, we measure the extent to which we can accurately measure the salience of topics/concepts that might be of interest to an analyst tasked with analyzing the content posted on social media platforms. We also evaluate whether concepts like positive and negative sentiment can be meaningfully extracted from Social Media content. As a test case, we examined Twitter content generated by female users who are sympathetic to the Islamic State in Iraq and Syria (ISIS). Although the results were based on a small sample of users, we demonstrate that ISIS fangirls differ in the content of their tweets from other, non-radicalized, teenage girls, and that automated text analysis techniques can detect the differences. The basic technique proposed here is a promising step in devising techniques for quantifying the salient topics being discussed on social media platforms, and should be developed further to create more fine-grained exanimations of such content.