EUVIMEDOpdag EUVIMED
Systematische ÜbersichtOpen access · CC BY-NC 4.0

AI for multilingual mental health recognition via social media analysis:A systematic review

Alfiyah Qothrunnada Zulfa, Sunu Wibirama, Syukron Abu Ishaq Alfarozi, Sri Kusrohmaniah

Array · 2026-12 · 32 · 101373

DOI: 10.1016/j.array.2026.101373 ↗

At a glance

Systematische Übersicht in Array, 2026-12. Written by Alfiyah Qothrunnada Zulfa and 3 co-authors. 156 references. The full text is freely available (open access, CC BY-NC 4.0). Funded by: Ministry of Higher Education Science and Technology Republic of Indonesia.

AbstractBibliografiske oplysningerCiteRelated studies

Abstract (original language)

Mental disorders affect approximately one billion individuals worldwide. Although conventional methods have the potential to diagnose these conditions, they suffer from societal stigma, insufficient facilities, and prolonged procedures. Social media provides an alternative approach that enables users to openly express their health issues, facilitating the use of artificial intelligence (AI) for early recognition. However, users’ communication styles vary owing to cultural and language differences. To address this problem, we present a systematic review of 94 articles from 2018 to 2025 that implemented AI to identify mental health-related issues through social media, focusing on studies analyzing languages other than English. Our review indicates that research has concentrated on three mental health conditions: depression, suicidal ideation, and anxiety, while conditions such as stress, bipolar, Obsessive-Compulsive Disorder (OCD), anorexia, Post-Traumatic Stress Disorder (PTSD), and Attention Deficit Hyperactivity Disorder (ADHD) remain underexplored. The top three languages in research are Chinese ( n =21), Arabic ( n =20), Thai ( n =12), and Bengali ( n =12), with additional studies in Indonesian, Spanish, Korean, Roman Urdu, Filipino, Russian, Sinhala, Brazilian Portuguese, German, Dutch, Japanese, French, and Malay. The methodologies have progressed from machine learning (ML) to deep learning (DL) approaches and transformer-based models (BERT, RoBERTa, DistilBERT). Challenges include data sparsity, class imbalance, annotation noise, culture-specificity, and the trade-off between the scalability of labeling based on data collection and label quality. Future research should include multilingual datasets, more attention to underexplored areas of mental health conditions, explainable AI, and domain-specific language models.

Funding

  • Ministry of Higher Education Science and Technology Republic of Indonesia

Cite

APA

Zulfa, A. Q., Wibirama, S., Alfarozi, S. A. I., & Kusrohmaniah, S. (2026). AI for multilingual mental health recognition via social media analysis:A systematic review. Array, 32, 101373. https://doi.org/10.1016/j.array.2026.101373

Vancouver

Zulfa AQ, Wibirama S, Alfarozi SAI, Kusrohmaniah S. AI for multilingual mental health recognition via social media analysis:A systematic review. Array. 2026;32:101373. doi:10.1016/j.array.2026.101373

BibTeX
@article{Zulfa2026,
  author = {Zulfa, Alfiyah Qothrunnada and Wibirama, Sunu and Alfarozi, Syukron Abu Ishaq and Kusrohmaniah, Sri},
  title = {AI for multilingual mental health recognition via social media analysis:A systematic review},
  journal = {Array},
  year = {2026},
  volume = {32},
  pages = {101373},
  doi = {10.1016/j.array.2026.101373}
}
Start for free →