Free beta access. 50 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

Here's that study.

H. Akın Ünver 22, Department of International Relations Özyeğin University İstanbul, Turkey, and Ahmet Kurnaz, Department of Public Administration and Political Science, Çanakkale Onsekiz Mart University, Çanakkale, Turkey, February 21st, 2022, “Securitization of Disinformation in NATO Lexicon: A Computational Text Analysis,” https://ssrn.com/abstract=4040148 [ML]

In order to explore how NATO has securitized disinformation in recent years, we have scraped 238,452 tweets from NATO and official affiliated accounts from January 2014 to February 2021 and extracted more than 2000 speeches, press releases, reviews, official texts, archival documents and publications from the NATO e-library (?). Out of this sample, we have extracted documents and content that contained the keywords ‘disinformation’, ‘misinformation’, ‘fake news’, ‘propaganda’, ‘hybrid warfare’ and ‘information warfare’, and logged the number of their occurrences within these texts by date. Since this study doesn’t focus on the semantic differences between these keywords and consider them as different references to disinformation as a discursive strategy, we code and merge them singularly as variable ‘disinfo’. Our preliminary analysis shows that NATO has used these keywords most frequently in tweets, followed by speeches, reviews and publications. However, when analyzed proportionally, NATO publications focus on disinformation most frequently (75% of all documents), followed by tweets (40%), reviews (36.97%) and official texts (21.05%). In this study, we employ structural topic modeling (STM) - a text analysis approach that finds ‘topics’ in an unstructured corpus based on covariate information (Roberts et al., 2019). It follows a statistical logic that measures co-occurrence likelihoods of keywords and terms that are more likely to appear with each other and derives topical meanings out of those likelihoods. Topic modeling is increasingly used in the study of large volumes of text in social sciences such as large archival documents, or social media text datasets by producing “each word on the basis of some number of preceding words or word classes,” and “generate[ing] words based on latent topic variables inferred from word correlations independent of the order in which the words appear” (Wallach, 2006). In recent years topic modeling has become a widely-used method to study large Twitter datasets, and political discussions that happen on other social media platforms (Giles, 2012; Hong, Davison, 2010; Liang, Fu, 2015).

Sign in to recut, update, or save this card.