Published: 2024

Prompt Patterns for Structured Data Extraction from Unstructured Text

CATEGORIES

RISK-BASED PROCESS SAFETY ELEMENTS

Research Summary

Introduces prompt patterns aimed at extracting structured records from unstructured text and discusses how to reduce common failure modes such as missing fields, inconsistent formatting, and ambiguity. This is directly applicable to PSM because many important inputs are narrative or semi-structured (incident/near‑miss reports, audit findings, operating procedures, and lessons learned). The patterns can help convert these documents into consistent, machine-readable outputs—e.g., event timelines, causal factors, barrier/safeguard lists, and action-item registers—which supports trending, verification of required elements, and improved handoff into PSM databases and workflows.

AUTHORS

Max Moundas; Jules White; Douglas C. Schmidt

CITATIONS

M. Moundas, J. White, and D. C. Schmidt, "Prompt Patterns for Structured Data Extraction from Unstructured Text," in Proc. 31st Conf. Pattern Languages of Programs, People, and Practices (PLoP 2024), Skamania Lodge, WA, USA, Oct. 2024, pp. 1–15.