Towards Identifying Fine-Grained Depression Symptoms from Memes

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 8890–8905 July 9-14, 2023 ©2023 Association for Computational Linguistics Towards Identifying Fine-Grained Depression Symptoms from Memes Shweta Yadav∗, Cornelia Caragea∗, Chenye Zhao∗, Naincy Kumari†, Marvin Solberg⋆, and Tanmay Sharma‡ ∗Department of Computer Science, University of Illinois Chicago, USA †Central University of Rajasthan, India ⋆Wayne State University, USA ‡Indian Institute of Technology Gandhinagar, India ∗{shwetay,cornelia,czhao43}@uic.edu, †knaincy818@gmail.com, ⋆marvin.solberg@wayne.edu, ‡tanmay.sharma@iitgn.ac.in Abstract The past decade has observed significant at- tention toward developing computational meth- ods for classifying social media data based on the presence or absence of mental health con- ditions. In the context of mental health, for clinicians to make an accurate diagnosis or pro- vide personalized intervention, it is crucial to identify fine-grained mental health symptoms. To this end, we conduct a focused study on de- pression disorder and introduce a new task of identifying fine-grained depressive symptoms from memes. Toward this, we create a high- quality dataset (RESTORE) annotated with 8 fine-grained depression symptoms based on the clinically adopted PHQ-9 questionnaire. We benchmark RESTORE on 20 strong monomodal and multimodal methods. Additionally, we show how imposing orthogonal constraints on textual and visual feature representations in a multimodal setting can enforce the model to learn non-redundant and de-correlated features leading to a better prediction of fine-grained depression symptoms. Further, we conduct an extensive human analysis and elaborate on the limitations of existing multimodal models that often overlook the implicit connection between visual and textual elements of a meme. 1 Introduction Mental health disorders have a profound impact on society. Almost 1 billion people worldwide suf- fer from mental health disorders, predominantly depression, anxiety, mood, and substance use disor- ders (WHO, 2022). Two of the most common men- tal health disorder, depression and anxiety account for the US $1 trillion in economic losses worldwide annually (Health, 2020). This cost is projected to rise to a staggering US $6 trillion by 2030 (Bloom et al., 2012). Apart from the economic burden, the social burden of mental health disorders is huge. Suicide is now the fourth leading cause of death among those aged 15 to 29 years old (Fleischmann Figure 1: Example of a depressive meme. If we merely evaluate the textual content (“These would give me a peaceful scene in a land of trouble”), it is difficult to establish the author’s true feelings. However, the meme- image can provide complementary information to help recognize the depression symptom (self-harm) correctly. et al., 2021). However, considering its preponder- ance and global burden, depression continues to be significantly undertreated in all practice settings, where fewer than one-third of adults with depres- sion receive effective treatment. Denial of illness and stigma are the two most common obstacles to appropriate diagnosis and treatment of depression (Sirey et al., 2001). Recently, social media data has emerged as a powerful “lens” for tracking and detecting depres- sion (De Choudhury et al., 2013; Yates et al., 2017). The vast majority of the existing works on …

Content truncated to 500 words