Identifying Key Topics Bearing Negative Sentiment on Twitter: Insights Concerning the 2015-2016 Zika Epidemic

"[U]understanding the characteristics of negative sentiment could help inform federal disease control agencies' efforts to disseminate relevant information to the public about Zika-related issues."

The purpose of this study was to analyse public sentiment concerning Zika using posts on Twitter and determine the qualitative characteristics of positive, negative, and neutral sentiments expressed. The researchers used all tweets, including personal communication as well as news articles, because news articles can go viral and include negative sentiment, as seen with a British Broadcasting Corporation (BBC) article stating that the United States (US) declared the Zika virus scarier than first thought.

The researchers chose to focus in particular on negative sentiment tweets, as this is what health officials will be most concerned with due to the greater need for intervention and information dissemination on these topics. For example, during the Ebola outbreak, it was found that failure to engage communities had detrimental effects, whereas engaging communities - e.g., through using varied methods of communication, organising regular meetings with the community, and identifying female and male community leaders to spread key messages - helped curtail the outbreak. Learnings from the Ebola and other experiences inspired public health officials in the US Centers for Disease Control and Prevention (CDC) held a live chat with the public and posted information on social media as they gained new information concerning the Zika virus. The nature of the new symptoms associated with Zika could have led to fear and anxiety among the public. Therefore, public health officials needed to continue to disseminate preventative methods and information on how to address symptoms to help mitigate possible panic. By understanding what is of concern to the public, officials can focus on targeting their messages to addressing these concerns.

This study draws on data obtained in a previous study using a semantic Web application that aids comprehension of social perceptions by semantics-based processing of massive volumes of event-centric data on social media. In the previous study, tweets were collected between February 24 2016 and April 27 2016, using the keywords "Zika", "Zika virus", and "Zika virus treatment". For the present study, 2 researchers annotated 5,303 randomly selected tweets from a total of 48,734 tweets. A supervised machine learning classifier was built to classify tweets into 3 sentiment categories: positive, neutral, and negative. Tweets in each category were then examined using a topic-modeling approach to determine the main topics for each category, with focus on the negative category.

The majority of tweets displayed negative sentiment (2,423; 46% of the total tweets) and the fewest displayed positive sentiment (1,010; 19%). The total number of negative tweets was almost 4 times larger than the positive and neutral categories combined. Selected findings:

• Within the positive sentiment themes, there were 4 broad qualitative topics: mosquito-killing methods, models to help understand the Zika virus, detection of the Zika virus in cells, and treatment and prevention discoveries (e.g., the development of vaccines to treat Zika). These topics reflect positive public perception because, for example, knowing where Zika accumulates would help with developing treatments. Tweets in the positive category used words with positive connotations such as "understand", "develop", "hope", "discover", "benefit", and "reveal". The broader themes were labeled based on domain expertise and from journals such as Vaccine and Morbidity and Mortality Weekly Report (MMWR), allowing further categorisation of 10 topics.
• Themes in the neutral category mostly comprised posts from public health experts and news agencies informing the public and thus are more likely to state facts than opinions. There were 3 broad qualitative themes: public health messages (e.g., explanations of how scientists were trying to unravel the Zika mystery), knowledge gaps, and Zika characteristics.
• In the negative sentiment topics, there were 3 broad topics: neural defects caused by Zika infection, abnormalities because of Zika infection, and reports and findings concerning the Zika virus (e.g., the BBC article that describes Zika as scarier than initially though).

This analysis suggests that, as the majority of topics in the negative sentiment category concerned symptoms, officials should focus on spreading information about prevention and treatment research. (Specific messaging suggestions are provided for each of the negative sentiment categories.) Deficiencies in communication among the media, the public, politicians, and scientists can heighten public concern. For example, when nonexperts express views different from experts, public fear can increase. The researchers suggest that experts need to continue disseminating factual information and also keep peer reviewing each other to make sure studies such as the one by Andrew Wakefield et al. in The Lancet incorrectly suggesting vaccines cause autism do not get published in the future. Furthermore, scientists tend to use words the public does not understand, such as the word "asymptomatic", causing a discrepancy between what is stated by public health officials and what the reader interprets. This can be addressed by scientists better explaining their work at an elementary school level.