Detecting depression stigma on social media: A linguistic analysis

Area of Science:

• Computational linguistics
• Social media analysis
• Mental health stigma research

Background:

• Detecting depression stigma in mass media is vital for effective intervention strategies.
• Chinese social media platforms like Sina Weibo are significant sources of public discourse on mental health.

Purpose of the Study:

• To develop computational models for identifying depression stigma expressions in Chinese social media posts.
• To analyze the prevalence and types of depression stigma present in these posts.

Main Methods:

• Collected and analyzed 15,879 Sina Weibo posts using keyword searches.
• Performed content analysis to classify posts as containing depression stigma or not.
• Built classification models using four algorithms (Simple Logistic Regression, Multilayer Perceptron Neural Networks, Support Vector Machine, Random Forest) based on linguistic features.

Main Results:

• 6.09% of posts (967 out of 15,879) indicated depression stigma.
• Common stigmatizing views included "People with depression are unpredictable" (39.30%), "Depression is a sign of personal weakness" (15.82%), and "Depression is not a real medical illness" (14.99%).
• Achieved highest F-Measure values of 75.2% for stigma vs. non-stigma detection and 86.2% for classifying specific stigma types.

Conclusions:

• Linguistic analysis methods enhance the online detection of depression stigma.
• These computational models can improve the performance of stigma reduction programs.
• Findings highlight the potential of AI in understanding and combating mental health stigma in digital spaces.