JMIR Publications

ABSTRACT

Background:

Depression becomes a serious personal and public mental health problem nowadays. However, it is not easy to discover patients with depression as they cannot disclose or discuss their mental health conditions with others easily. Most work is based on self-report, which is time-consuming, and usually miss a certain quantity of cases. Therefore, automatic discovering patients with depression from other sources, such as social media, attracts more and more attention. Social media as one of the most important daily communication systems connect a large quantities of people including depression patients, and provides one channel to discover depression patients. In this paper, we investigate deep learning methods on depression risk prediction for Chinese microblogs, which has potential for discovering patients with depression and tracing their mental health conditions.

Objective:

The aim of this study is to explore the performance of some state-of-the-art deep learning methods on depression risk prediction for Chinese microblogs.

Methods:

Deep learning methods with pretrained language representation models, including BERT, RoBERTa and XLENT, are investigated and evaluated on an annotated depression benchmark dataset collected from Weibo for depression risk prediction. We also compare the different deep learning methods in two settings: 1) using publicly released pretrained language representation models directly, and 2) further pretraining language representation models from 1) on a large-scale unlabeled dataset collected from Weibo. Precision, recall and F1 score are performance evaluation measures.

Results:

The deep learning methods achieves the best macro F1 score of 0.547, a new benchmark result over the depression benchmark dataset. Specially, the deep learning methods achieves the highest macro precision of 0.536, recall of 0.376 and F1 score of 0.420 on microblogs of depression risk.

Conclusions:

In this study, we use deep learning methods with pretrained language representation models to predict depression risk for Chinese microblogs automatically. The experimental results show that the deep learning methods perform better than previous methods and have greater potential to discover patients with depression and trace their mental health conditions.