Extracting Topics from Unstructured Text via LLM Function Composition: Measuring Partisan Coverage Dynamics in American Television News, 2000-2025

• In Event: Text as Data

Brief Overview

I formalize an oracle‑preserving, merge‑robust LLM summarization and labelling pipeline, release an accompanying Python package, and apply it to US television news to measure partisan and topical change from 2000 to 2025.

Author

• Mitchell Linegar, Caltech