Paper Type
Complete
Paper Number
PACIS2025-1436
Description
Depression is a major contributor to the global disease burden, with depressive tendencies requiring early detection and intervention. This study proposes a Symptom Recognition Framework based on Cognitive Graph of Depression Tendency Symptoms (DTS-KG-SRM), rooted in schema theory. It systematically integrates domain-specific and general knowledge to construct a cognitive graph of depressive symptoms. Entities and relationships are extracted using the ChatGPT API and designed prompts. Leveraging graph activation mechanisms and advanced feature extraction, the framework combines textual and graph features for precise, multi-label classification of depressive symptoms in user posts. Tests on the self constructed DTSD dataset show DTS-KG-SRM outperforms baseline models like BERT and Mental RoBERTa. Ablation studies confirm each module’s importance. This study innovatively designs a recognition framework, optimizes cognitive graph construction, enhances knowledge integration, and delivers superior performance, addressing a research gap and providing technical support for early depression diagnosis.
Recommended Citation
Deng, Shasha and Ji, Xin, "Fine-Grained Recognition of Depressive Tendency Symptoms Based on Cognitive Graphs" (2025). PACIS 2025 Proceedings. 16.
https://aisel.aisnet.org/pacis2025/ishealthcare/ishealthcare/16
Fine-Grained Recognition of Depressive Tendency Symptoms Based on Cognitive Graphs
Depression is a major contributor to the global disease burden, with depressive tendencies requiring early detection and intervention. This study proposes a Symptom Recognition Framework based on Cognitive Graph of Depression Tendency Symptoms (DTS-KG-SRM), rooted in schema theory. It systematically integrates domain-specific and general knowledge to construct a cognitive graph of depressive symptoms. Entities and relationships are extracted using the ChatGPT API and designed prompts. Leveraging graph activation mechanisms and advanced feature extraction, the framework combines textual and graph features for precise, multi-label classification of depressive symptoms in user posts. Tests on the self constructed DTSD dataset show DTS-KG-SRM outperforms baseline models like BERT and Mental RoBERTa. Ablation studies confirm each module’s importance. This study innovatively designs a recognition framework, optimizes cognitive graph construction, enhances knowledge integration, and delivers superior performance, addressing a research gap and providing technical support for early depression diagnosis.
Comments
Healthcare