Heure et date : le 6 mars à 12h30
Lieu : PK-4610 et Zoom
Conférencière : Raqeebir Rab (Ahsanullah University of Science and Technology)
Titre : Personality and Emotion – A Comprehensive Analysis Using Contextual Text Embeddings (SBERT)

NB : La présentation sera donnée par Zoom. Nous vous invitons à nous joindre au local PK-4610 pour un visionnement en groupe à partir de 12h30. La pizza sera fournie à partir de 12h. 

Inscriptions : Veuillez vous inscrire en indiquant votre prénom, nom et choix de garniture de pizza préférée à ce lien.

Abstract : Personality and emotions have always been closely intertwined since humans have evolved to adapt to these two forms. Emotions are indicative of a person's personality and vice versa. This paper aims to investigate the complex relationship between these two fundamental aspects of human behavior. The objective is to automate the process of determining the relationship between personality traits of the MBTI (Myers-Briggs Type Indicator) and Ekman emotions based on the context of user-written social media posts using contextual embedding (SBERT). A robust mechanism is employed to achieve this, involving two distinct phases to identify emotions. The first phase of the research entails quantifying the correlation between predetermined emotions and social media posts by utilizing cosine similarity scores. Phase 2 introduces a cross-dataset learning approach where several machine learning models are trained on various datasets to identify emotions in textual content. After training, these models utilize the patterns they learned to predict emotions in a target dataset. Support Vector Machine (SVM) is chosen among all the trained models based on performance and effectiveness in emotion prediction tasks. We developed a vetting mechanism combining two approaches to improve accuracy, reliability, and trustworthiness for the final emotion prediction. This paper finds patterns that link MBTI personalities with Ekman emotions. It showcases a detailed display of emotion distributions within each MBTI personality type. This comprehensive study provides valuable information on how people with different types of personalities typically express emotions on social media.

Biographie :  Raqeebir Rab received her B.Sc. degree in science with a major in computer science from the Augustana Faculty, University of Alberta, Canada in 2004. She completed her M.Sc. in computer science at the Concordia University, Montreal, Canada in 2012. She is an Assistant Professor at the Department of Computer Science and Engineering (CSE), Ahsanuallah University of Engineering and Technology (AUST), Dhaka, Bangladesh. Her research interests include wireless multihop networks (ad hoc and sensor networks) with an emphasis on mathematical modeling, performance analysis, protocol design, and data science.