As the explainability of mental disorder detection models has become important, symptom-based methods that predict disorders from identified symptoms have been widely utilized. However, since these approaches focused on the presence of symptoms, the context of symptoms can be often ignored, leading to missing important contextual information related to detecting mental disorders. Furthermore, the result of disorder detection can be vulnerable to errors that may occur in identifying symptoms. To address these issues, we propose a novel framework that detects mental disorders by leveraging symptoms and their context while mitigating potential errors in symptom identification. In this way, we propose to use large language models to effectively extract contextual information and introduce an uncertainty-aware decision fusion network that combines predictions of multiple models based on quantified uncertainty values. To evaluate the proposed method, we constructed a new Korean mental health dataset annotated by experts, named KoMOS. Experimental results demonstrate that the proposed model accurately detects mental disorders even in situations where symptom information is incomplete.

Early intervention for suicide risks with social media data has increasingly received great attention. Using a suicide dictionary created by mental health experts is one of the effective ways to detect suicidal ideation. However, little attention has been paid to validate whether and how the existing dictionaries for other languages (i.e., English and Chinese) can be used for predicting suicidal ideation for a low-resource language (i.e., Korean) where a knowledge-based suicide dictionary has not yet been developed. To this end, we propose a cross-lingual suicidal ideation detection model that can identify whether a given social media post includes suicidal ideation or not. To utilize the existing suicide dictionaries developed for other languages (i.e., English and Chinese) in word embedding, our model translates a post written in the target language (i.e., Korean) into English and Chinese, and then uses the separate suicidal-oriented word embeddings developed for English and Chinese, respectively. By applying an ensemble approach for different languages, the model achieves high accuracy, over 87%. We believe our model is useful in accessing suicidal ideation using social media data for preventing potential suicide risk in an early stage.