Multimodal Depression Detection Model fusing Emotion Knowledge Graph
Zikai Wang, Bin Deng, Xin Yi Shu, Jian Shu · 2023
More and more patients with depression use social platforms to share their daily life. The abundant multimodal information contained in Weibo makes the early detection of depression possible. This paper proposes a multimodal depression detection model fusing emotion knowledge graph(EKG-MDDM). To construct user text sequences, EKG-MDDM combines emotional knowledge graph vocabulary and BS algorithm to process user blog posts. EKG-MDDM uses ALBERT and AGG16 networks to extract text and image features. In addition, this paper summarizes and extracts 11 behavioral features from user’s descriptions and posting behaviors. Finally, the fusion features are applied to the task of depression detection. This model effectively exploits the emotional information contained in user blog posts. Compared with the previous recognition methods, the proposed model is superior to the baseline model in terms of Accuracy, Precision, Recall and F1 indicators.