Hierarchical Explainable Network for Investigating Depression From Multilingual Textual Data

Abstract

Depression is the most prevalent mental health disorder, which is spreading like an epidemic. With the rise of the internet, people from all over the world now express their daily lives on social media, which can be utilized to analyze their mental health condition. Automated approaches like Machine Learning (ML) and Deep Learning (DL) are used to analyze social media posts. The ML techniques are performing well in detecting depression from monolingual text. Nevertheless, ML model’s low interpretability makes it difficult to identify depression in multilingual texts. Hence, the study focuses on seven languages for developing the model. Traditional ML and DL algorithms like Convolutional Neural Network (CNN), Support Vector Machine, Random Forest, and Decision Tree lack interpretability regarding prediction. To effectively analyze the multilingual dataset, CNN with a Hierarchical Attention Network is implemented in the model and then the Gaussian Naive Bayes algorithm is applied to the extracted output layer to generate an interpretable prediction, allowing it to offer enhanced explainability compared to other state-of-the-art algorithms. The proposed approach demonstrates effectiveness in analyzing multilingual texts with an accuracy of 89.09% outperforming other ML and DL models in terms of time and memory efficiency.

Keywords

• Convolutional neural network
• depression detection
• explainability
• hierarchical attention network
• multilingual