According to WHO data, mental health is a major source of concern throughout the world, amounting to suicide in the majority of instances if left untreated. Presently, social media is a great way for people to express themselves through text, emoticons, images, or videos that depict their sentiments and emotions. This has opened up the possibility of investigating social networks in order to better comprehend their users' mental states. Social media is currently being used by researchers to predict the prevalence of mental illnesses such as depression, suicide ideation, anxiety, and stress. This area of study has considerable potential for the monitoring, diagnosis, and prevention of mental health issues. In this research we aim to give an overview of the most recent works for predicting mental health status on social media. We focused on data collection and annotation approaches, preprocessing and feature selection, model selection, and validation. We addressed research on depression and suicidal thoughts, which are the most common among mental health studies on social media.
Top1. Introduction
Mental illness is a significant public health concern on a global scale due to the suffering, morbidity, dysfunction, and financial trouble caused by it. Estimates from the WHO and the National Mental Health Survey 2017 show about 197.3 million people in India need professional care for several mental health issues. This comprised around 45.7 million people with depressive illnesses and 45 million people with anxiety disorders. The issue has been compounded by the Covid-19 outbreak, making it a global concern (The Indian Express, 2022). Although there are various types of mental health problems in India, the most common among them is depression (World Health Organization, n.d.). Lancet investigations show a substantial relationship between mental health diseases and the suicide mortality rate in India (Statistica, 2022). Mental health determinants include the skill to control thoughts, emotions, behaviors, and interactions with others. Along with particular psychological, behavioral, and genetic characteristics, environmental, social, cultural, economic, and political factors also impact the mental health of an individual(Sagar et al., 2020). Effective therapy for mental illness depends on an early diagnosis (Zimmerman&Coryell, 1987). However, a significant number of people experiencing depressive symptoms still choose not to seek professional care as a result of the societal stigma attached to mental illness (De Choudhury et al., 2013). Thus, individuals frequently rely on social media to handle their concerns.
Social networking services are popular destinations for people to connect and share their emotions and thoughts. People suffering from mental health concerns found solace in online forums, tweets, and blogs, where they could express their thoughts and experiences implicitly or overtly (Tsugawa et al., 2015; Reece & Danforth, 2017). Social media enables the analysis of online data for assessing user emotions and moods to understand their behavior during social media use. People suffering from any form of mental illness have a tendency to behave in a different manner on social media, producing enough data for modeling different characteristics. Mental health studies show a strong correlation between a person's language use and their emotional health. Research on data from social media started around 2013 and many mental health issues such as depression (Tsugawa et al, 2015; De Choudhury et al., 2013; Reece et al., 2017; Ríssola et al., 2020; Samanta et al., 2023; Sarkar et al., 2022; Park et al., 2012; Bathina et al., 2021; Nadeem, 2016; Preoţiuc-Pietro et al., 2015; Maupomé & Meurs, 2018; Schwartz et al., 2014; Resnik et al., 2015; Tsugawa et al., 2015; Nguyen et al., 2014; Wolohan et al., 2018; Tyshchenko, 2018; Tadesse et al., 2019; Benamara et al., 2018; Song et al., 2018; Cacheda et al., 2019; Shen et al., 2017; Lora et al., 2020; Stankevich et al., 2020), suicide ideation (Sarkar et al., 2023; Coppersmith et al., 2016; O'dea et al., 2015; Burnap et al., 2015; Shing et al., 2018; Chatterjee et al., 2022a; Wongkoblap et al., 2017; Coppersmith et al., 2015; Dholariya, 2017; Pandey et al., 2019; Nock et al., 2008; Bilsen,2018; Klonsky et al., 2016; Fernandes et al., 2018; Nock et al., 2012; Ji et al., 2020; Cook et al., 2016; Mbarek et al., 2019; Ji et al., 2018; Ramírez-Cifuentes et al., 2020; Na et al., 2020; Duberstein et al., 2000; Islam et al., 2018; Al Asad et al., 2019; Chatterjee et al., 2021; Angskun et al., 2022; Gupta et al., 2022; Cha et al., 2022; Shen et al., 2017), schizophrenia (Mitchellet al., 2015;Hänsel et al., 2021; Bae et al., 2021), eating disorders (Chancellor et al., 2016; Wang et al., 2017; Paul et al., 2018; Kumar et al., 2019) and others have been assessed with a high accuracy of 80-90%. In addition, these methods can also be used to measure related disorders, for instance, stress, and self-harm without the need for an in-person assessment. Using this notion, several researchers have developed new types of future interventions and techniques for identifying mental health illnesses in their early stages. To do this, Natural Language Processing (NLP) approaches are used along with machine learning techniques to spot psychological disorders in user submissions.