Home » WAPOR 2017 Conference

WAPOR 70th Annual Conference

15th-17th July 2017

Lisbon, Portugal

Back to Programme

A Trilingual and Comparative Approach to Understanding the Conversation about the Zika Virus on Social Media

Dominique Brossard (University of Wisconsin-Madison)
Jennifer Chung (University of Wisconsin-Madison)
Dominique Brossard (University of Wisconsin-Madison)
Dietram Scheufele (University of Wisconsin-Madison)
Michael Xenos (University of Wisconsin-Madison)
Luisa Massarani (Red de popularización de la Ciencia y la Tecnología en América latina y el caribe)
Andrew Maynard (Arizona State University)

Keywords: Social media, big data, sentiment analysis, and emerging technologies

Abstract

Social media is often described as a way to globalize communication. Platforms, like Facebook and Twitter, provide a way for people to share opinions with others from around the world, without geographic barriers. While social media has the potential to eliminate geographic limitations, it does not remove language barriers. Differences in language create several sub-conversations, with potentially meaningful variation in the sentiments and opinions being expressed. These variations in social media conversations are especially important for issues with global policy-making implications, like the recent outbreak of the Zika virus. Over the course of 2016, the virus quickly spread across the Western Hemisphere, infecting people in countries that speak different languages and have very different media systems. This study takes a multilingual and comparative approach to better understanding the conversations and opinions being expressed about the Zika virus on Facebook and Twitter. We are specifically interested in how the online discussions vary between languages over time. To better understand the broader online opinion climate, we looked at three different facets of the conversation surrounding Zika: if posts assigned blame, if a problem or solution frame was used, and what strategies to fight the virus were being discussed. We used the machine learning platform Crimson Hexagon ForSight to collect and analyze the data. The ForSight platform combines machine learning with human coding. For the human coding, we ran separate series of inter-coder reliability trials for each codebook, in each language, and for both Facebook and Twitter. Our preliminary results show meaningful differences in sentiments and opinions expressed in each language. For Twitter specifically, we found differences in what strategies to fight the outbreak in all three languages. There were also a wide variety in the volume of posts in each language, with far less Tweets in Portuguese than in English and Spanish. These meaningful differences would have been lost, had we not studied all three languages. This research emphasizes there is not only one online conversation and set of opinions about global policy issues, but many language-based conversations and opinions being expressed. In addition to this analysis, we are also interested in examining the different media systems that these conversations come from. To do so, we looked at a sub-sample of posts that contain country-level geographic information. Then looked at factors like internet access, the adaption rates and demographics for different media types, and the relationships between traditional media and government for the countries of interest. We then compared these media system variables with the conversations about Zika in each country. This research gives insight into the different media and opinion forming climates surrounding a global health policy issue.


© 2015 WAPOR All Rights Reserved