Work

A data driven framework for mental health states assessment in social platforms

Year: 2022
Type: dissertation
Abstract: The link between mental disorders and social media usage has led researchers to work on the development of automated methods to detect mental health issues in social platforms. This thesis proposes a framework for mental health states assessment, considering suicidal ideation, eating disorders, depression and alcoholism as use cases. This framework is composed of modules dedicated to the characterization and detection of mental disorders, along with the definition of a non-invasive support-provision approach based on social recommendation. We make use of user characterization techniques, and propose several predictive models based on behavioral and multimodal data. We also propose a contact recommendation approach, evaluated by people with anorexia nervosa, which prioritizes the recommendation of harmless accounts to follow in social platforms. The main contributions of this work are: 1) insights regarding the behavior of users with mental disorders in social media; 2) methods for the enhancement of text representations (word embeddings) adapted to binary and multiclass predictive tasks that address domain specific tasks, and that handle small data; 3) several predictive models for the detection of mental disorders; and 4) the definition and evaluation of a contact recommendation method dedicated to users with anorexia. This method has been proven to be helpful for counteracting the over-personalization effects caused by social platforms’ recommender systems. This research work is focused on the analysis of data from Spanish and English speakers, addressing multiple social platforms. With the outcomes of this thesis we expect to contribute to the further development of tools to assist experts and help users living with mental disorders.
Language: English
Field: Psychology
Open Access status: green
OpenAlex ID: W7008953749
Cites: 88
Cited by:
Related to: 10