How to Detect Emotions in Text

Wang, Y., & Pal, A. (2015). Detecting emotions in social media: A constrained optimization approach. In Proceedings of the 24th International Joint Conference on Artificial Intelligence (pp. 996–1002). Retrieved from https://www.ijcai.org/Proceedings/15 /Papers/145.pdf

In this paper, the authors propose a “constraint optimization framework” to identify emotions within social media content. Such data can be used to understand user behaviors and preferences and improve user interactions. The authors designed the framework specifically to address emotion detection challenges including the presence of multiple emotions, topic correlation, emotion bindings, and noisy labels. The authors detail in depth the formulas used for each step in the process to show the construction of the framework. To test the framework, the authors used three datasets: SemEval, ISEAR, and a Twitter dataset of 1800 tweets. After testing with different models both quantitatively and through qualitative evaluation, the results showed the framework was effective in detecting emotion from text in a linear fashion suitable for large datasets. The framework is configurable to add new features and incorporate refined emotion lexicons, and solves multi-label classification problems with the capacity to allow for multiple emotions, and assign and identify emotion categories and their dominance.

‍