The surge in internet use for expression of personal thoughts and beliefs has made it increasingly feasible for the social Natural Language Processing (NLP) research community to find and validate associations between
social media posts
and
mental health status
. Cross-sectional and longitudinal studies of low-resourced social media data bring to fore the importance of real-time responsible Artificial Intelligence (AI) models for mental health analysis in native languages. Aiming to classify research for social computing and tracking advances in the development of learning-based models, we propose a comprehensive survey on
mental health analysis for social media
and posit the need of analyzing
low-resourced social media data for mental health
. We first classify three components for computing on social media as:
SM
- data mining/ natural language processing on
social media
,
IA
-
integrated applications
with social media data and user-network modeling, and
NM
- user and
network modeling
on social networks. To this end, we posit the need of mental health analysis in different languages of East Asia (e.g. Chinese, Japanese, Korean), South Asia (Hindi, Bengali, Tamil), Southeast Asia (Malay, Thai, Vietnamese), European languages (Spanish, French) and the Middle East (Arabic). Our comprehensive study examines available resources and recent advances in low-resourced languages for different aspects of SM, IA and NM to discover new frontiers as potential field of research.