Analyzing the words people use in written (or typed) text, especially after a negative event, is a long-standing practice in psychological research (Tausczik & Pennebaker, 2010). Several researchers have capitalized on this methodology and employed analyses of large corpora of tweets and their content to study emotional responses to collective traumas in communities across the world. Most studies rely on dictionaries from the Linguistic Inquiry and Word Count program (LIWC; Pennebaker, Both, Boyd, & Francis, 2015) to do so. This dictionary approach captures the content of tweets based on empirically validated lists of words that possess psychological meaning (e.g., anxiety). Although there are other natural language processing techniques (see Jones, Brymer, & Silver, 2019 for some alternatives), researchers have used the LIWC dictionary approach because it is effective and easy to implement with Twitter data.
Studying large-scale traumatic events is difficult, given the unpredictability of such events and the resources needed to mobilize in their aftermath to conduct sound research (Jones, Wojcik, Sweeting, & Silver, 2016; Silver, 2004). However, big social media data on Twitter offer a multitude of ways to study people in stressful contexts that avoid many of the practical and methodological challenges associated with this work. These data can be readily accessed by researchers with a bit of technical know-how and can provide a time-scale of individual-level emotion expression before and after large-scale traumatic events, not a common feature of traditionally collected data. Importantly, posts on Twitter (i.e., tweets) contain text data that represent emotional states, intentions for behaviors, thoughts and activities of its users.