Classifies text into structured semantic units with authority, risk, and attention scores. Enables deterministic preprocessing for AI agents to filter and route content without using an LLM.
