New chapter: From Source to Structure — Extracting Knowledge Graphs with LLMs

Malte Vogl | Mar 31, 2026 min read
Process diagram of Stage 1 of the extraction pipeline: preprocessing, chunking, LLM extraction and validation, sampling and evaluation, with a human expert reviewing.

Extracting structured knowledge graphs from unstructured historical texts is precise but slow when done by hand, and hard to automate without losing the nuances of historical sources. This chapter presents a two-stage computational pipeline — open information extraction followed by LLM-based validation — built to connect large, previously isolated sources like biographical lexicons into usable knowledge graphs.

Read the paper

  • Authors: Raphael Schlattmann, Aleksandra Kaye, and Malte Vogl
  • In: Understanding Science with Large Language Models? Potentials for the History, Philosophy, and Sociology of Science, ed. Arno Simons, Adrian Wüthrich, Michael Zichert, and Gerd Graßhoff (transcript, 2026), pp. 307–334
  • DOI: 10.14361/9783839447529-307
  • Featured image: detail from the accompanying poster Silo to Structure (Stage 1 of the extraction pipeline)