Liking, sharing, commenting and reacting on Facebook: User behaviors' impact on sentiment intensity

The form of communication on Facebook is not only limited to posting and commenting, but also includes sharing, liking and reacting. This study looks into how a Facebook diabetes community uses like, comment, share and reaction in expressing themselves online and how these distinctions can be used to improve sentiment classification from text extracted from the said group. An intensity formula using those behaviors was proposed and experimentations conducted using Weka. The findings reveal a model encompassing user behaviors is able to determine sentiment more accurately compared to one without, with a 94.6 percentage of accuracy. Additional analyses reveal behaviors such as liking, commenting and sharing to contribute more to the sentiment classification compared to reacting. This further cement the need to include such behavioral aspects into sentiment polarity calculation, as it would help algorithms achieve better predictability when classifying sentiment.