Multi-Aspect Transfer Learning for Detecting Low Resource Mental Disorders on Social Media

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 3202–3219 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 3202 Multi-Aspect Transfer Learning for Detecting Low Resource Mental Disorders on Social Media Ana Sabina Uban1,3, Berta Chulvi1,2, Paolo Rosso1 1Universitat Polit`ecnica de Val`encia, Val`encia, Spain, 2Universitat de Val`encia, Valencia, Spain, 3University of Bucharest, Bucharest, Romania auban@fmi.unibuc.ro, berta.chulvi@upv.es, prosso@dsic.upv.es Abstract Mental disorders are a serious and increasingly relevant public health issue. NLP methods have the potential to assist with automatic mental health disorder detection, but building annotated datasets for this task can be challenging; moreover, annotated data is very scarce for disorders other than depression. Understanding the commonalities between certain disorders is also important for clinicians who face the problem of shifting standards of diagnosis. We propose that transfer learning with linguistic features can be useful for approaching both the technical problem of improving mental disorder detection in the context of data scarcity, and the clinical problem of understanding the overlapping symptoms between certain disorders. In this paper, we target four disorders: depression, PTSD, anorexia and self-harm. We explore multi-aspect transfer learning for detecting mental disorders from social media texts, using deep learning models with multi-aspect representations of language (including multiple types of interpretable linguistic features). We explore different transfer learning strategies for cross-disorder and cross-platform transfer, and show that transfer learning can be effective for improving prediction performance for disorders where little annotated data is available. We offer insights into which lin- guistic features are the most useful vehicles for transferring knowledge, through ablation experiments, as well as error analysis. Keywords: Mental disorders, depression, transfer learning, explainability, social media, low resource 1. Introduction Mental health disorders are an important and pervasive public health issue. Depression in particular affects approximately 300 million people worldwide (World Health Organization, 2012), and the problem has re- cently been exhacerbated by the COVID-19 pandemic (Lima et al., 2020; Shah et al., 2020; Shigemura et al., 2020; Torales et al., 2020; Xiang et al., 2020). Men- tal health disorders are closely connected to suicide, with depression present in 35% of suicides (Bridge et al., 2006). Self-harm is a strong predictor of suicide, present in the recent histories of around 40% of sui- cides (Cavanagh et al., 2003). Moreover, mental dis- orders are massively underdiagnosed and undertreated (Sheehan, 2004; Allan et al., 2014), with more than half of the people suffering from depression not receiv- ing any treatment. People affected by mental disor- ders are often reluctant to approach a specialized clin- ician to seek help with treating the disorder. However, more and more frequently people turn to social me- dia to discuss their issues and to seek emotional sup- port. This opens up an important opportunity for auto- matic processing of social media data in order to iden- tify changes in mental health status that may otherwise go undetected before they develop more serious health consequences. The way mental disorders manifest and can be recog- nized is primarily …

Content truncated to 500 words