Elsevier and LG AI Research Automate Chemistry Image Extraction

Chemical drawings encode critical data through bonds, atoms, and spatial relationships that traditional text-based indexing fails to capture. By integrating LG AI Research’s proprietary vision technology, Elsevier can now automatically parse images from journals and patents, allowing chemists to bypass manual document review. This transition from static visual files to indexed knowledge enables faster novelty searches and more reliable synthesis planning.

The new model combines molecule detection with optical chemical structure recognition, outperforming existing benchmarks in accuracy and scale. According to Mirit Eldor, Managing Director of Life Sciences at Elsevier, this automation aims to reclaim hours previously lost to deciphering figures, turning dead-end graphics into actionable data. Hwayoung Edward Lee of LG AI Research noted that the system was engineered specifically to interpret the nuances of chemical layouts with human-expert precision.

Following the successful implementation of substance extraction, the partnership is shifting focus toward reaction parsing. This next phase aims to map complex chemical transformations, further expanding the dataset available to researchers. All processes remain governed by Elsevier’s existing ethical frameworks, ensuring that automated insights maintain the rigorous quality standards required for scientific and clinical decision-making.