FedTherapist: Mental Health Monitoring with User-Generated Linguistic Expressions on Smartphones via Federated Learning

• DOI: 10.18653/v1/2023.emnlp-main.734
• Publication Date: 2023-12-10T21:58:19Z
• Authors: Jaemin Shin; Hyungjun Yoon; Seungjoo Lee; Sungjoon Park; Yunxin Liu; Jinho Choi; Sung-Ju Lee
• Keywords: FOS: Computer and information sciences; Computer Science - Machine Learning; Computer Science - Computation and Language; Artificial Intelligence (cs.AI); Computer Science - Artificial Intelligence; Computation and Language (cs.CL); 3. Good health; Machine Learning (cs.LG)

Psychiatrists diagnose mental disorders via the linguistic use of patients. Still, due to data privacy, existing passive health monitoring systems alternative features such as activity, app usage, and location mobile devices. We propose FedTherapist, a system that utilizes continuous speech keyboard input in privacy-preserving way federated learning. explore multiple model designs by comparing their performance overhead for FedTherapist overcome complex nature on-device language training on smartphones. further Context-Aware Language Learning (CALL) methodology effectively utilize smartphones' large noisy text signal sensing. Our IRB-approved evaluation prediction self-reported depression, stress, anxiety, mood from 46 participants shows higher accuracy compared with non-language features, achieving 0.15 AUROC improvement 8.21% MAE reduction.