Automatic inference of affective charge in social media posts

• Grant number: 21/07067-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: September 01, 2021
• End date: December 31, 2022
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Helena de Medeiros Caseli
• Grantee: Fernanda Malheiros Assi
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil
• Associated scholarship(s): 22/02472-6 - Development of an affective load inference module for tweets in politics, BE.EP.IC

Abstract

Social media are increasingly becoming spaces where people share their opinions, emotion, and perceptions about various subjects including politics, culture, and personal issues. In this context, many studies and researches on natural language processing (NLP) have been focusing on text processing in social media in order to, among other reasons, foresee political ideologies, personality disorders, and mental health. However, in order to find such traits, the majority of these researches focus on detecting polarity (valence) or emotions between documents or between users of the social media, instead of analyzing texts across time. Thus, this project aims to create a computational model capable of measuring the emotional charge over time, in social media posts, linked to a specific topic or event through a time series of emotional domains. The valence and the average excitation estimated from posts published within the same small window of time will represent an observation of such a time series, which describes a variation of the emotional charge in the "Scherer's space". In order to predict the emotional charge, it is intended to use the LSTM model and GRU cells as base models, due to the fact that in recent years they have proven to be promising models with their capability to model long-term dependencies. (AU)