Self-harm and suicidal behaviors are preventable if the risk can be identified early. It is critical to incorporate external knowledge to enhance the suicide prediction performance. In this talk, I will introduce our recent research on developing knowledge-aware suicide risk prediction models in different contexts. First, we propose a novel patient embedding method, namely Dx2Vec (Diagnoses to Vector), based on the comorbidity network constructed by all historical diagnoses. Dx2Vec represents the comorbidity patterns among diseases and temporal patterns of historical admissions for each patient. Experiments with territory-wide clinical data demonstrate that the Dx2Vec-based model outperforms the baseline deep learning model in identifying patients who would self-harm/suicide within 12 months. Second, we develop a suicide knowledge graph, based on which we propose a knowledgeaware Natural Language Processing (NLP) model to identify the help-seekers at a high suicide risk by mining their counselling conversations. Experiments with the data of a major Cantonese mental health counselling platform demonstrate that the proposed model significantly outperformed state-of-the-art NLP models by incorporating domain knowledge. The findings have the potential to be translated into effective self-harm/suicide prevention programs..

Speaker: Dr Qingpeng ZHANG
Date:  9 September 2020 (Wed)
Time: 14:00pm - 15:00pm
Poster: Click here