Original texts and analysis code are not provided(contact me if needed).
All labels are in Korean.
- Embedding Model: pre-trained SBERTs, OpenAI embedding models (models can be changed depending on the setting)
- Clustering: HDBSCAN with UMAP
- Topic Representation: News headline-based OpenAI summarization(gpt4.1-mini, model choice may vary)
- Resolve bookmarks issue