Caracterização e previsão do tom emocional de usuários das comunidades online de transtornos mentais

• Ano de defesa: 2019
• Autor(a) principal: Bárbara Silveira Fraga
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ICX - DEPARTAMENTO DE CIÊNCIA DA COMPUTAÇÃO; Programa de Pós-Graduação em Ciência da Computação; UFMG
• Palavras-chave em Português: Computação; Análise de sentimento; Saúde Mental; Aprendizado do Computador; Saúde Mental - Teses; Aprendizado do Computador - Teses; Computação - Teses; Redes sociais on-line - Teses.; Análise de sentimento - Teses
• Link de acesso: http://hdl.handle.net/1843/32052

The alarming increase of the number of people affected by mental health disorders has become one of the major public health problems faced by governments around the world. Traditional clinical interventions require high budgets and might not include a considerable number of people struggling to improve their mental health conditions. Thus, several studies have investigated alternative interventions approaches capable of reach further the target population and enable continuous interaction at lower costs. One such alternative is using online social networks (OSN), which in recent years have connected people willing to exchange experiences related to health problems (e.g., obesity and depression). The main objective of this dissertation is to analyze how communities focused on the discussion of mental health disorders are able to help and improve their user’s health conditions . Through characterizing users from four Reddit communities (Depression, SuicideWatch, Anxiety, and Bipolar), we analyzed how interactions through posts and comments might influence the emotional tone (state) of these users. In other words, we analyzed whether seeking help on these networks is effective and if it’s followed by behavioral changes in how users express their feelings over time. Our results show that the user’s emotional tone changes over time,which 68% of cases are considered a positive change, providing evidence that support and encouragement among users in these communities are effective to improve their patients’ mental health. participants. Additionally, we propose predictive models to capture the variation of users’ emotional tone. This variation takes on values in [-2,2]. Our models accurately capture the variation of users’ emotional state: in the worst case, the mean square error value was 1.062 and in the best case 0.5747. The models presented might assist interventions promoted by health professionals in social networks in order to support people with mental health disorders.