Emotional Trends in Social Media: A State Space Approach

Sorry, this entry is only available in European Spanish. For the sake of viewer convenience, the content is shown below in the alternative language. You may click the link to switch the active language.

“In this paper, a state space system is used to model a po- litical discussion held on Twitter platform. This approach is based on the assumption of psychology research – humans react emotionally to events. Therefore, certain features of conversations are modeled into state space system, which deals with events and its transitions. Emotions and other additional features from Twitter are used for model learning. The focus of this paper is set on the modeling of the emotional features of a Twitter discussion. The proposed approach is further evaluated with an example discussion about the Spanish cor- ruption affair held on Twitter during summer 2013. The experimen- tal results show a method to make hidden emotional trends apparent, also within short trending topics.”

El anterior es el abstract del artículo presentado por Sören Volgmann junto conmigo, Oliver Niggermann y Paolo Rosso donde describe una aproximación de espacio de estados (ver figura) y filtros Kalman para detectar tendencias emocionales en Social Media. Básicamente, se modela la conversación en torno a un tema (caso Bárcenas) en función de la evolución de las emociones (6 emociones básicas de Ekman) y la pasión, se discretiza en diferentes unidades de tiempo, se aprenden los parámetros de un espacio de estados, se aplica un filtro Kalman para separar la estacionalidad y el ruido para manifestar más claramente la tendencia.

No cuento los resultados, hay que esperar a su aceptación, o en su defecto, intentarlo de nuevo en otro congreso 🙂