Deep Learning Architectures and Strategies for Early Detection of Self-harm and Depression Level Prediction
Year: 2020
Type: article
Abstract: [EN] This paper summarizes the contributions of the PRHLT-\nUPV team as a participant in the eRisk 2020 tasks on self-harm detection and prediction of depression levels from social media. Computational methods based on machine learning and natural language processing have a great potential to assist with early detection of mental disorders of social media users, based on their online activity.We use multi-dimensional representations of language, and compare various deep learning models' performance, exploring rarely approached avenues in previous research, including hierarchical deep learning architectures and pre-trained transformers and language models.
Authors: Ana Sabina Uban, Paolo Rosso
Language: English
Topic: Mental Health via Writing
Subfield: Social Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
Funder: Generalitat Valenciana
OpenAlex ID: W3094646185