Academic Research

Automated text analysis methods have made possible the classification of large corpora of text by measures such as topic and tone. Here, we provide a guide to help researchers navigate the consequential decisions they need to make before any measure can be produced from the text. We consider, both theoretically and empirically, the effects of such choices using as a running example efforts to measure the tone of New York Times coverage of the economy. We show that two reasonable approaches to corpus selection yield radically different corpora and we advocate for the use of keyword searches rather than predefined subject categories provided by news archives. We demonstrate the benefits of coding using article segments instead of sentences as units of analysis. We show that, given a fixed number of codings, it is better to increase the number of unique documents coded rather than the number of coders for each document. Finally, we find that supervised machine learning algorithms outperform dictionaries on a number of criteria. Overall, we intend this guide to serve as a reminder to analysts that thoughtfulness and human validation are key to text-as-data methods, particularly in an age when it is all too easy to computationally classify texts without attending to the methodological choices therein.

Computational propaganda and the use of automated accounts in social media have recently become the focus of public attention, with alleged Russian government activities abroad provoking particularly widespread interest. However, even in the Russian domestic context, where anecdotal evidence of state activity online goes back almost a decade, no public systematic attempt has been made to dissect the population of Russian social media bots by their political orientation. We address this gap by developing a deep neural network classifier that separates pro-regime, anti-regime, and neutral Russian Twitter bots. Our method relies on supervised machine learning and a new large set of labeled accounts, rather than externally obtained account affiliations or orientation of elites. We also illustrate the use of our method by applying it to bots operating in Russian political Twitter from 2015 to 2017 and show that both pro- and anti-Kremlin bots had a substantial presence on Twitter.

Automated and semiautomated Twitter accounts, bots, have recently gained significant public attention due to their potential interference in the political realm. In this study, we develop a methodology for detecting bots on Twitter using an ensemble of classifiers and apply it to study bot activity within political discussions in the Russian Twittersphere. We focus on the interval from February 2014 to December 2015, an especially consequential period in Russian politics. Among accounts actively Tweeting about Russian politics, we find that on the majority of days, the proportion of Tweets produced by bots exceeds 50%. We reveal bot characteristics that distinguish them from humans in this corpus, and find that the software platform used for Tweeting is among the best predictors of bots. Finally, we find suggestive evidence that one prominent activity that bots were involved in on Russian political Twitter is the spread of news stories and promotion of media who produce them.

This chapter examines the use of social networking sites such as Twitter in measuring public opinion. It first considers the opportunities and challenges that are involved in conducting public opinion surveys using social media data. Three challenges are discussed: identifying political opinion, representativeness of social media users, and aggregating from individual responses to public opinion. The chapter outlines some of the strategies for overcoming these challenges and proceeds by highlighting some of the novel uses for social media that have fewer direct analogs in traditional survey work. Finally, it suggests new directions for a research agenda in using social media for public opinion work.