Interpretable Unsupervised Community Detection with LLM-Symbolized Structured Processes | AIChainDay