Automatic Information Extraction from Palaeontological Literature Using R

Automatic Information Extraction from Palaeontological Literature Using R

Palaeontological data extraction

Guoyi Zhao

Daniel Cashmore

Assessing fossil record completeness requires researchers to manually extract detailed anatomical and specimen information from scientific literature, making large-scale studies time-consuming. This project develops PALEAN, an R-based workflow for automatically converting unstructured palaeontological literature into structured research data. The pipeline reads scientific papers using text extraction and OCR, cleans and searches the resulting text, identifies relevant taxonomic, anatomical, locality and specimen information, and exports structured tables for further analysis. The workflow was tested on three recent theropod dinosaur papers describing Aratasaurus museunacionali, Thanatotheristes degrootorum and Tralkasaurus cuyi. Preliminary manual verification shows that the pipeline can recover key taxonomic information, specimen identifiers, skeletal elements and formation-level locality information, although complex anatomical descriptions and some locality details still require manual review. The project demonstrates the potential for reproducible automated workflows to reduce the manual effort involved in fossil completeness data collection.

palaeontology, fossil record, fossil completeness, data extraction, text mining, R, automation, theropod dinosaurs, natural language processing