Analyzing Education-Related Posts in Detecting Depression on Social Media Using Topic Modeling Method and Large Language Model

• Published in: 2025 13th International Conference on Information and Education Technology (ICIET) pp. 355 - 359
• Main Authors: Thamrin, Syauki A., Chen, Arbee L.P.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 18.04.2025
• Subjects: Analytical models; Data models; Deep learning; Depression; Depression detection; Education; Large language models; Mental disorders; Social media; Social networking (online); Stress; Topic modeling; Transformers
• DOI: 10.1109/ICIET66371.2025.11046300

Depression is a mental disorder that negatively affects people's daily life. The causes of depression are different for different people, and aspects of life affected by depression are also different. In the education field, academic pressure affects the performance of the student in school and possibly becomes the cause of depression. People with depression tend to hide their feelings or are not even aware they are depressed, causing them to be difficult to recognize. Since some people with depression tend to express their feelings on social media, deep learning models were used to detect depression from social media posts. However, the explanation for the classification result is limited to negativemeaning words in the posts. Therefore, key factors related to depression, such as due to education, are difficult to understand. Our study aims to analyze education-related posts and their relations to other posts posted by the same depression users. We utilized a transformer model to identify education-related posts, and the topics of the posts were obtained using the topic modeling method and large language model. By analyzing the posts posted before and after the education-related posts, our proposed method can identify characteristics that differentiate depression and non-depression users. Based on our experiment and analyses, topics related to social media interactions are identified more in depression users than non-depression users. This result may indicate that social media interactions happen as a process of coping with the academic pressure for the depression users.