Using text mining and sentiment analysis to analyse YouTube Italian videos concerning vaccination

Abstract

Background: Recently, social networks have become a popular source of information on health topics. Particularly, in Italy, there is a lively discussion on the web regarding vaccines also because there is low vaccination coverage, vaccines hesitancy, and anti-vaccine movements. For these reasons, in 2017, Institutions have introduced a law to force children to make ten compulsory vaccines for school attendance and proposed a vaccination campaign. On social networks, this law has fostered a fierce discussion between pro-vaccinations and anti-vaccinations people. This paper aims to understand if and how the population's opinion has changed before the law and after the vaccination campaign using the titles of the videos uploaded on Youtube in these periods. Method: Using co-occurrence network (CON) and sentiment analysis, we analysed the topics of YouTube Italian videos on vaccines in 2017 and 2018. Results: The CON confirms that vaccinations were very disapproved before the law. Instead, after the communication campaign, people start to be less critical. The sentiment analysis shows that the intense vaccination campaign also promoted by medical doctors pushed the sentiment to change polarity from a prevailing negative opinion in 2017 (52% negative) to a positive one in 2018 (54% positive). Conclusion: At the population level, the potential misinformation of social networks could be significant and is a real risk for health. Our study highlights that vaccination campaigns on social networks could be an essential instrument of health policies and a sharp weapon to fight ignorance and misrepresentations of non-qualified people influencing individuals' decision-making.

Multi-perspectives systematic review on the applications of sentiment analysis for vaccine hesitancy

“…In addition, other aspects in this class could be attributed to the high level of data noise [ 66 , 85 ], including spelling mistakes and the use of non-standard text [ 73 ]. Finally, complex views on the same topic [ 89 ], a variety of online content [ 90 ], objectivity and subjectivity [ 89 ], the structure of data [ 87 ], or even their uneven distribution could pose significant challenges in determining the nature of the data [ 68 ].…”

“…The other part of this subgroup included papers discussing sentiment analysis with relation to vaccine hesitancy for public opinions. The work by [ 90 ] discussed the adoption of text mining and sentiment analysis to analyze Italian YouTube videos concerning vaccination. The authors used co-occurrence network (CON) and sentiment analysis to analyze the topics of these videos from May 1 to October 1 for years 2017 and 2018.…”

Characterizing HIV discussions and engagement on Twitter

“…By aggregating and analyzing data from various sources (e.g., social media, search engines, mobile phones, location services, crowdsourcing, and wearable devices), digital surveillance complements traditional public health surveillance systems [3][4][5]. Novel potentials of digital data (such as real-time, organic, and volume) from location logs, web searches, and online reviews have supported in tracking, predicting, and preventing diseases [6,7] as well as other critical public health concerns such as substance abuse [8], public attitudes towards vaccination [9], and self-injurious thoughts and behaviors [10]. Similarly, social media has recently emerged as a significant vein of digital surveillance systems.…”

Prospective associations of regional social media messages with attitudes and actual vaccination: A big data and survey study of the influenza vaccine in the United States

“…A growing number of important studies have analyzed vaccine content on Facebook, Twitter, and YouTube. Although some studies have involved manual coding of posts [7] , [8] , [9] , others have used computational techniques such as co-occurrence network analysis [10] , and yet others have combined manual coding with computational methods, such as a support vector machine [11] . A recent study of anti- and pro-vaccine clips on YouTube identified words such as “chemical,” “mercury,” and “toxic” in anti-vaccine content and “hospital,” “chronic,” and “unvaccinated” in pro-vaccine material [9] .…”