An Agentic Workflow for Detecting Personally Identifiable Information in Crash Narratives

Junyi Ma, Pei Li, Rui Gan, Kai Cheng, Steven T. Parker, Bin Ran · 2026

Read An Agentic Workflow for Detecting Personally Identifiable Information in Crash Narratives by Junyi Ma, Pei Li, Rui Gan, Kai Cheng, Steven T. Parker, Bin Ran (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-15369.txt.

Description

Ma and five colleagues build a locally run agentic workflow that finds personal details in crash narratives, where sparse and uneven names, addresses, and plate numbers block wider safety research. A hybrid extractor routes structured items to rules and tricky items to a fine-tuned model, with ensemble extraction and an evidence-based verifier filtering false hits. On real crash data it beats baselines, reaching precision 0.82, recall 0.94, and accuracy 0.96.

Catalog record

CREATORJunyi Ma, Pei Li, Rui Gan, Kai Cheng, Steven T. Parker, Bin Ran
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE45141
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.15369

Browse the full library · About the Tower of Babel