The extraction of molecular structures and reaction data from scientific documents is challenging due to their varied, unstructured chemical formats and complex document layouts. To address this, we introduce MolMole, a vision-based deep learning framework that unifies molecule detection, reaction diagram parsing, and optical chemical structure recognition (OCSR) into a single pipeline for automating the extraction of chemical data directly from page-level documents. Recognizing the lack of a standard page-level benchmark and evaluation metric, we also present a testset of 550 pages annotated with molecule bounding boxes, reaction labels, and MOLfiles, along with a novel evaluation metric. Experimental results demonstrate that MolMole outperforms existing toolkits on both our benchmark and public datasets. The benchmark testset will be publicly available, and the MolMole toolkit will be accessible soon through an interactive demo on the LG AI Research website.
@misc{research2025molmolemoleculeminingscientific,
title={MolMole: Molecule Mining from Scientific Literature},
author={LG AI Research and Sehyun Chun and Jiye Kim and Ahra Jo and Yeonsik Jo and Seungyul Oh and Seungjun Lee and Kwangrok Ryoo and Jongmin Lee and Seung Hwan Kim and Byung Jun Kang and Soonyoung Lee and Jun Ha Park and Chanwoo Moon and Jiwon Ham and Haein Lee and Heejae Han and Jaeseung Byun and Soojong Do and Minju Ha and Dongyun Kim and Kyunghoon Bae and Woohyung Lim and Edward Hwayoung Lee and Yongmin Park and Jeongsang Yu and Gerrard Jeongwon Jo and Yeonjung Hong and Kyungjae Yoo and Sehui Han and Jaewan Lee and Changyoung Park and Kijeong Jeon and Sihyuk Yi},
year={2025},
eprint={2505.03777},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2505.03777},
}