Googling YouTube for public awareness campaign -Stroke. Big data analytic approach.

Abstract

Background: It is important to educate the public to recognise the signs and symptoms of stroke as treatment are currently available. Social media (YouTube) has been used to educate the public on stroke signs. In this study we use machine learning approach to examine these campaigns against exemplar 2006 campaign in North America (Stroke Heroes Act Fast). Further we explore the sentiment in the North American campaign to evaluate social media responses. Method(s): The number of views and comments on these videos were extracted from YouTube application programming interface (API). Natural language processing steps consisted of stripping number, punctuations, stop words and parsing comments into separate sentences. Sentiment analysis of the words were performed using NRC word emotion association lexicon. Result(s): Over 5 campaigns (2009-2015), there were 14895 views (0.06360597% of the Australian population) and 1 comment on videos (xK1Qf0MTIRU [re-enactment of a stroke], RBaMgsSKzCc [Know the warning signs], MmoGeCXov [Join the Fight Stroke], 7pbdKLOHNU [Fire in the Brain] and xA-P5voEik8 [FAST Train]). The exemplar campaign from North America (YHzz2cXBlGk) has 406,253 views (0.12% of North American population) and 123 comments. Sentiment expressed include trust (25.7%), fear (22.4%), joy (20.2%), sadness (20.8%), surprise (7.1%), anticipation (21.3%), disgust (4.9%), anger (4.4%), negative (30.1%) and positive (44.8%). Conclusion(s): The Australian social media campaigns have not garnered as many views or comments. The overwhelming positive sentiment in the North American campaign provides clues for developing material on social media for education.