Constructing a mental health analysis system for social media using large language models

As social media platforms have penetrated every aspect of peoples lives, many mental health problems have also arisen along with them. In todays digital age, analyzing mental health trends through these platforms has become critical. In this study, we present a system designed to identify the mental health trends of Weibo users by extracting and analyzing the content posted by users on Weibo, Chinas leading social media platform. The system is mainly composed of two parts: a data acquisition module and an analysis module. The data collection module uses the Python-based web scraping tool Scrapy to scrape comments from popular topics on Weibo. At the heart of the analysis module is a large language model fine-tuned from a psychological database. The module assesses the topic and specific content of the posts, scoring comments based on criteria such as positivity, alignment with mood disorders, and potential signs of psychoactive substance use. This data is stored and mediated using the relational database MySQL, and then analyzed and visualized using advanced data analysis tools. Through this method, we can timely and comprehensively monitor the mental health status of social media platforms, and provide a solid foundation for further academic research on public mental health.