With the increase of Smartphone use, there is a growing need for advanced features that offer to Smartphone users a smarter interaction. We aim through the presented system to detect users' emotions from their textual exchanges, dealing with the complexity of chat writing style and the evolution of languages. We consider that such a system is a start for interesting applications that exploit users' emotional states. Our system uses an unsupervised machine learning algorithm that performs emotion classification, based on a data corpus built from YouTube comments. The reason behind such a choice is the similarity between YouTube comments and instant messages writing style. To classify a text entry into a particular emotion category, we compute its similarity to each target emotion, using the Pointwise Mutual Information measure. Our method yields a global precision of 92.75%, which reflects the feasibility of our approach.
Yasmina, D., Hajar, M., & Hassan, A. M. (2016). Using YouTube Comments for Text-based Emotion Recognition. In Procedia Computer Science (Vol. 83, pp. 292–299). Elsevier. https://doi.org/10.1016/j.procs.2016.04.128