SOLO: Socially Observed Linked Opinion
Social media, such as Facebook, Twitter, and YouTube, have provided researchers with a significant amount of data about human interaction. One of the most common analyses used on these large social media data sets is sentiment analysis typically used in reference to how individuals feel about a specific topic. However, the standard tools used for this analysis have three inherent problems: they are approaches developed to analyze large bodies of text, they ignore the social context of social media, and they do not consider the international dimension of social media. We have a new method for this type of assessment that addresses these problems by not relying on language modifiers, by considering the original post that individuals are responding to, and by formalizing observed interaction using affect control theory, which allows for cross cultural analysis.
This project will verify our approach by utilizing native English speakers to manually classify social media text along four dimensions taken from affect control theory: positive sentiment, negative sentiment, potency, and activity. By developing a gold standard, this project will demonstrate the validity of applying our method to social media texts.